Skip to main content

Full text of "The pronunciation of the Greek aspirates"

See other formats


«5. < 



fyxmll Hmvmitg pibai^g 




iienrg W. Sage 


,A-->g-c£.^.;?:-^ 'f/l/?.:^'.. 

_ Cornell University Library 

PA 267.D26 

3 1924 021 602 002 

The original of tliis book is in 
tine Cornell University Library. 

There are no known copyright restrictions in 
the United States on the use of the text. 



ELIZABETH A. S. DAWES, M.A., D.Lit.(Lond.) 


D. NUTT, 270-271 STRAND 





P. 12 1. 27, omit "and Curtius." 

P. 53 1. li, for "I" read "a." 

P. 62 1. 22, /o?' "ff<(> for air" read "ait for o-f." 

P. 79 1. 25, /OJ- "irup((/)ou = 7rup(x<'"" ''^"'^ " nuppi<|)ou = riuppix'"' " 

P. 98 1. 19, after "elisions" insert "from various dialects." 

P. 103 1. 6, for " of the two " read "of either of the two." 

those who maintain that, whether or not the Greek aspirates 
were originally genuine aspirates, they had already in the 
classical period become spirants, or at least affricatives. 
Amongst these we may reckon Arendt, Brlicke, Ebel, Telfy, 
Papademetrakopoulos, Raumer, Roscher, and Rumpelt. 

In this small treatise I purpose to set forth concisely the 
various arguments of different authorities, and to attempt to 
estimate from the evidence thus collected the possibility cf 
drawing conclusions cf a more or less certain nature. 

B 2 



In a former treatise on the Pronunciation of Greek I did 
not consider that of the aspirates; in the following pages I 
propose to discuss the latter within the bounds of a limited 

The authorities on the question of the pronunciation of the 
Greek aspirates may be roughly divided into two classes. 
First, -vye have those who support the theory that the Greek 
aspirates were pronounced as the Indian aspirates, that is to 
say, were genuine aspirates, even in the classical period of the 
Greek language. To this class belong Brugmann, G. Curtius 
the great champions of the Erasmian pronunciation, Blass, 
Meisterhans, G. Meyer, and many others. Secondly, there are 
those who maintain that, whether or not the Greek aspirates 
were originally genuine aspirates, they had already in the 
classical period become spirants, or at least affricatives. 
Amongst these we may reckon Arendt, Briicke, Ebel, Telfy, 
Papademetrakopoulos, Raumer, Roscher, and Rumpelt. 

In this small treatise I purpose to set forth concisely the 
various arguments of different authorities, and to attempt to 
estimate from the evidence thus collected the possibility cf 
drawing conclusions cf a more or less certain nature. 

B 2 


With this intention I would at the outset disclaim the bias 
of any foregone conclusion, and maintain that I have en- 
deavoured to collate evidence with strict impartiality. 

As regards the estimation of conclusions to be drawn from 
such evidence, I can only offer my suggestions with the greatest 
deference to the authority of the many learned scholars who 
have given their opinion on the subject under discussion. 

The question of the pronunciation of the Greek aspirates 
treated fully and in .detail would suffice to fill a volume of 
some size; here, therefore, I have limited myself to the 
consideration in detail of a part only of the material at my 
disposal. In choosing from the latter I have preferred to 
select for especial investigation the internal evidence sup- 
plied by Greek, as I consider it is from this source that 
we can draw the most satisfactory conclusions. The fact 
alone that, in dealing with internal evidence, we are not so 
dependent on reasoning from analogy, as we are when treating 
of the external evidence to be derived from other sources, 
enables us to base our arguments on more certain ground. 

The method on which I have proceeded has been to investi- 
gate what degree of continuity is traceable in the successive 
periods of the language. 

I would claim to have especially elaborated that part which 
bears on the interpretation of the ancient grammarians, as also 
to have worked out more fully than has been done before the 
history of the interchange of letters, and the evolution of the 
phonetic laws, as illustrated by modern Greek. 

Although the investigation of Greek evidence thus con- 
stitutes the main substance of this inquiry, I have prefaced 
this by a short inquiry into the Indo-German sounds which the 
Greek aspirates are held to represent. 

Missing Page 


Briicke. — Grundzilge d. Physiologie. 1876. 

Psichari. — Mim. de la Soc. Zing, de Paris, vi. p. 304. 

Papademetrakopoulos. — ^dcravo^ tmp -irepl -rrj'; eXkr/p. rrpo- 

(pofja.'i .... 1889. 

C. Foy. — Lautsystem d. griech. Vulgarsprache. 

M. Deffner. — Zakonischs Gram. 

W. Prellwitz. — Etymol. Worterhuch der griech^. Sprache, 1.BiQ2, 
and many others. 

Chap. I. 

/Nature of the change | Have we a parallel in any 

other languaffe? 



If we suppose an intermediate stage we have 
a parallel in German. 


if the change was effected directly, we know 
of no parallel in any one language, but 
can compare the triimsitioii of 

/Period at which it may /Different /as to its completion /not in lirst centuries a.d. (Curtius, 

have taken place. 


!//as to its prelude 

Blass, &c.' 

in Classical period (including Homeric) 
'' (Arendt, Papademotrakopoidos;). 

/Boeotian 6 and <f> spirants in iifth century B.e. 
(Ahrena, Meister, Zacher). 

Laeonian and Cretan 9 also early spirant 
(Ahrens, Meister, Brugmann). 

/"perhaps in the ancient dialects (Blass). 

/Indo-German aspirated tenues into Gorman surd spirants 

/'Indo-German aspirated medial into Latin /* and, perhaps,/". 


/Sextus Empii'icus, second 
century a.d., places 
them with fifji[<^iava 

may mean that they were already 
spii'ants in his time. 

^.Priscian about 500 a.d. | difficult to distinguish between <^ and f 

Byz. Schol. of 
Dion. Thrax 
between 500 
/ .and 700 a.d. 

describes a spirantic pronoun (Blass). 

Manner in which it may have 
/ been effected 

.^ Direct transition (Ax-endt). 

by moans of a Transitional 
/ Stage 

Ch.ap. II. 
4.ffricativo Theory. 

/Explanation of /according to Raumer and Roscher 
/according to G. Meyer. 

Application of -to the Classical Period (Raumer and 

/to no definite period (G. Meyer). 

/Arguments pro /Such writings as : okxo<;, Sairc^ti). 

0, </), X' sometimes count as double 


letters in metre. 

Argviments con. /Such writings as ok\o^, ifcc, possibly 
due to poetic licence for the sake 
of the metro. 

/6, <(), x> generally single letters in metre. 

^ the grammarians never call these 
letters StTrAa. 

impossible to pronounce (f)9, x^ '^^ 


consecutive affricatives. 

Double homo- 
/■ Aspirate. 

/difficult to explain according to aspi- 
ratic theory. 

shows last stage in development of 
/ spirant. 



According to the generally accepted theory xi'^ were true 
surd aspirates, i.e. tenues + breathing in the classical period of 
Greek literature, remained such down to the first centuries of 
the Christian era, and then gradually became surd spirants, such 
as we find them in the present language (either by passing 
through the intermediate step of " affricatae," or without passing 
through any such transitional stage). Thus a great change 
must have taken place in the sound of these letters, and it may 
justly be asked whether a change of such a nature can be 
paralleled from any other language, as having taken place 
within somewhat the same compass of time. On investigation, 
if we suppose an intermediate stage of aifricatae,we find a parallel 
exists in German in those words where the Modern German 
spirants and affricatae which were caused by the consonantal 
change in High-German, have developed out of primitive German 
tenues, apparently by the following process : 

tenues : aspirates : affricatae : spirants 

(ef, Paul, Grundriss d. germ. Fhilog. p. 294) : 
and the different steps can, in many cases, be illustrated by 
examples from existent dialects, thus for this change at the 
beginnifig of words we have : 


Goth, (or N. Franc.) 

M. Franc. 

Ordinary' Germ, 



pfund or fund 

and Franc. 

Ord. Germ. 



Alem. Xann 

N. Franc. 

Ord. Germ. 


zu ( = tzu) 

and in the interior of words : 

Goth, hilpan O.G. helpfan Mod. Germ, helfen 
Goth, vitan O.G. wizzan Mod. Germ, wissen 

Goth, vakan O.G. wahhen Mod. Germ, wachen 
(where O.G. s is a voiceless spirant and A = p^;). 

The tenues of (Gothic) primitive Gerntian became aspirated 
(just as the Germans of to-day aspirate their tenues), and would 
then be parallel to the Greek y(^, <j), if pronounced originally as 
aspirates ; the affricative stage of the German words can be com- 
pared with such writings as oK'^pt;, %aTr<f>a), and finally the Modern 
German spirants, e.g. helfen, with Modern Greek spirants, eg. 
^vcn<i ( = fysis). Perhaps one may go further and say the 
dialectical ' punt ' and ' kan,' with persistence of original tenuis, 
may be compared with Modern Asiatic ep/co/iai, a-TOKa^m, which 
are also taken as instances of the survival of an original tenuis. 

It may be objected that this is not a complete parallel, 
because in German this change is by no means fully accomplished 
as yet, when initial, for in the dental series it has only gone as 
far as the affricative z, and in the guttural the affricative and 
spirants are both still wanting in ordinary German, and even in the 
labial series the pf is not yet universally spoken as simple /. 

But then German is, as far as our knowledge of it extends, a 
much younger language than Greek, as the Gothic we know is 


that of the fourth century A.D., so that Greek, of which we 
kuow a little from the seventh or sixth century B.C., is about 
ten centuries older, and, as the change of aspirates into spirants 
had already begun in the Greek dialects before the Christian 
era, there has been comparatively a far longer space of time for 
this change to become complete in Greek. 

Indeed in Modern German we see the change in process and 
as yet far from complete initially, whereas in Modern Greek not 
only has the change completely come, but it is in some cases 
going again — that is, the spirants of the middle ages are already 
undergoing a species of deterioration, if it may be so termed ; 
instead of only just becoming firmly fixed they are under certain 
conditions disappearing again, as if they had existed already 
many centuries and were now undergoing certain modifications — 
thus in the case of two consecutive spirants, the second one is 
now generally changed into a tenuis — e.^r. ^Odvm becomes 

On the other hand, if we consider with Arendt that the inter- 
mediate stage of affricatae need not necessarily have existed, but 
that the aspirated tenues changed directly into spirants (per- 
haps too without ever having been spoken in Greek for any 
length of time as aspirated tenues), we cannot compare this 
change with a similar change in one and the same language, but 
only with a change like the possibly direct transition of the 
Indo-German (pure and) aspirated tenues into German surd 

spirants, e.g. 

Skt. phena O.G. feim 

Skt. ^'tn Goth, freis 

and also that of the Indo-German aspirated mediae into Latin Ji, 
when initial, and perhaps /, for we are not quite sure whether / 
was an affricative or spirant. We think, therefore, that the change 


suggested for Greek is paralleled by changes that have and are 
still taking place in other languages, whether we consider that 
it was effected directly or by the help of an intermediate stage. 

With regard to the period by which this change had been 
completed, opinions also differ widely. 

Curtius, Blass, Zacher, and Peile think it was not completed 
by the first centuries A.D. 

Curtius (ii. p. I7)adds ," that the Laconians seem to have 
started it," and Blass (p. 108, Eng. ed.) that " this later pro- 
nunciation will not have arisen all at once, it must have needed 
time to have made its way from the lower to the upper stratum 
of the people, and to have become general. But its prelude is 
perhaps already to be found in the ancient Greek dialects." He 
further concludes that Laconian of fourth and third centuries 
B.C. had a partly spirantic 6 — that " in Cretan as we know it 
from the Gortynian inscription," the dental aspirate had become 
a spirant, but ^ and ^(^ had not ; and with regard to Boeotian, 
Lokrian, and Elean he maintains nothing. 

Zacher (p. 39) says with regard to the date of the change that 
" there must have been an approximation to the spirantic pro- 
nunciation at an early period, or special characters would not 
have been invented to represent the aspirates. 6 must have led 
the way as the sign @ is found in the oldest alphabets, and also 
it is never represented by TH." He maintains, however, that 
the spirantic pronunciation did not become general before the 
third century A.D., that is to say, in the written language, because 
in the dialects the change must partly have come about much 
earlier {e.g. the Spartan 6). 

Meisterhans {Oram. d. att. Inschr. § 27 ff.) says that the inter- 
change of tenues with % ^ ^ " excludes the possibility of the latter 
having been spirants in the classical period." Of ^ he says that 


"a real assimilation with the spirant / can be shown after 
120 A.D.," and of 6 that "the date of its change into a spirant 
cannot be fixed." 

Brugmann (Grimdriss, i. pp. 365-6) says that "in most 
dialects, e.g. in lon.-Att., % <^ ^ probably remained real aspirates 
down to the historic period. It is impossible definitely to fix 
the dates of their transition into spirants in dififerent localities, 
as we have no written evidence of a sufficiently conclusive 
nature to guide us." He proceeds to show that 6 became a 
spirant in Cretan, Laconian, Boeotian, Elean and Locrian at an 
early period, and that in Boeotian ^ also became a spirant 
very early. 

G. Meyer {Gk. Gram., pp. 213-214) likewise says that the date 
of the change cannot be even approximately determined. 

Meister (i. p. 260) and Ahrens say d and ^ were spirants in 
the fifth century B.C. inBoeotia, and Meister adds (ii. p. 54) that 
6 was a spirant in Crete before the fifth century B.C., and in 
Elis by the time of the Damokrates bronze, which according 
to Kirchhofif dates from about 300 B.C. 

Raumer concludes from Plato, CratyL, p. 427, that <^ pre- 
ceded X ^'^•i ^ i"^ becoming a spirant, and was already one in 
Plato's time. 

Briicke {GrundHi, pp. 127 &c.) maintains that in the combina- 
tions of aspirates, x^^ must always have been spirants, and 
probably also in all other cases, or else perhaps " affricatae." 
" It is doubtful," he says, " whether even in the very earliest 
ages of Greek speech and writing, the Greek aspirates were ever 
pronounced as tenues with a breathing appended. The pos- 
sibility of Greek aspirates having once been tenues -|- h cannot 
be denied, but inscriptions, change of tenuis into aspirate before 
the spiritus asper, &c., seem to give no support to this idea." 


Arendt supposes ')(^0 to have always been spirants in Greek; 

Psichari (Mem. de la Soc. de Ling., vi. p. 315) considers that 
X^6 were aspirates in classical times, and as regards the date of 
their change, he says : " Si Ton parvient a etablir la date du 
premier [v qui tombe devant x<j)0, on aura la date exacte du 
passage de I'aspiree k la spirante." 

Papademetrakopoulos (p. 621) argues that % ^ ^ were already 
spirants in the classical period, but allows that they may have 
been aspirates in the earliest stage of the language, that is, in 
the pre-Homeric times. 

We have indeed a few landmarks which indicate the time by 
which % ^ had for certain already become spirants, but they 
are not many and do not help towards proving when % ^ began 
to be spirants ; they are as follows : — 

(1) They were spirants by 700 (?) A.D., as shown by the de- 
scription of them given by the Scholiast on Dion. Thrax. 

(2) Priscian, 520 A.D., finds it difficult to distinguish between 
Latin/ and 0, and describes the latter as a labial spirant. 

(3) Sextus Empir., second century A.D., places them with 
rjfil<li(ova, which may mean that they were already spirants in 
his time. 

Thus, roughly speaking, ^^^ ^re unanimously considered » 
to have been spirants from the third . century A.D. onwards, 
and what remains to be examined is what they were before 
that time. 

As to the manner in which the transition from aspirated 
tenues to spirants was effected, some, like Arendt and Curtius, 
think that the Skt. sonant aspirates were in Greek immediately 
changed into surd aspirates, and then there was a direct transi- 
tion from them to spirants, though they differ as to the time 
when the second half of this change took place, for Arendt 


thinks that the surd aspirates probably never really obtained in 
Greek at all, whereas Curtius thinks they were spoken down to 
the first centuries *..D. 

Others, like Raumer, Roscher, and G. Meyer, think that in 
becoming spirants they passed through the transitional stage of 
" afifricatae," and further, the two former uphold the theory of 
X<l>0 having been real aflfricatae in the classical period. 

This theory, which may be called the afifricative, we will now 
proceed to discuss. 



The- manner of the development of spirants from aspirates 
through the stage of affricatae has been differently conceived by 
Raumer and G. Meyer. 

Raumer thinks that, first, the breathing of the aspirate 
became a spirant homogeneous to the tenuis, i.e. t + h became 
t + th, and then, secondly, the spirant crowded out the tenuis, 
i.e. t + th became "th": 

e.g. o(l>i<i = 0Tr<fii<i = 6<f)i<! (op + his = opfis = ofis). 

G. Meyer assumes the same first step, but the second one 
was, he says, that the tenuis became assimilated to the spirant' 
i.e. t + th became th + th, and then sometimes there was a third 
step by which the double spirant became single. 

Ba;)^o? = Ba«j^o? = Baj^j^o? and sometimes Baj^o?. 
bak + hos = bak + chos = Bachchos or Baclios. 


G. Meyer does not fix any definite period during which tlie 
pronunciation of ■x^cfyd as affricatae prevailed, whereas Raumer 
and Roscher think it was in use during the classical period. 
Blass thinks that, even if it did exist partially, it was not 
universal, but yet it will be as well shortly to review the 
arguments adduced in favour of it and those against it ; the 
latter being in our opinion quite sufiicient to condemn the 

The chief arguments for it are the words in which k^, t9, 
TT^ are evidently written for ■^, 6, cf) ; such are ta/c^j^jew, taA;;\;j; 
in the tragedians, from \/iaj^, and Ba/c^o? and "\aK'X^o<; ; also 
OKXO'i and oK'^eco found in the lyric poets. "ZaTrcfxo may 
perhaps be derived from ao^6<i ; from Vdhe we have correlative 
forms, such as ndrivr) and titOt), rirOevw, &c. Now these forms 
with K)(^ for T^ and trcf) for (e.g. crKVTr(f>o^ for a-KV(j>o<!, Hes.) 
are all poetical, and so written for the sake of the metre ; they 
are not the forms of ordinary use. It is impossible therefore to 
say whether these words were ever universally so pronounced 
or whether this orthography was merely a poetic licence, but 
most probably it was the latter, as this spelling is confined to a 
very few words. On an inscription we have the isola,ted foriA 
dvaTeTOeifiivoi<;, which may be another example in favour of i 
this theory, or may very likely only be a misspelling. ■ ■ • ■ 

The second argument is that % ^ ^ sometimes have the force 
of double letters in metre, and cause the lengthening of the 
preceding syllable : thus w'e have ^t\oa-o(f)ov in Aristoph. 
Mel. 571, ^po^ov in Theog., o^t? in Iliad and Hipponax, gene- 
rally Trl(f>avaKei,v in Homer (both l and I are Indo-Germ;), and 
KeKpv^aXov in Homer, but KeKpv(f>a\ov in classical Greek. These 
exceptions, however, are very few and do not occur in the 
tragedians, and may rightly be regarded as poetic licences, seeing 


that the general rule which holds good for all the poets is that 
% 0. ^ are treated as single betters. This consideration alone 
seems sufficient to refute the theory oi x'P^ being afifricatae 
{v. Blass, p. 103). 

Similarly it seems unreasonable to suppose that the gram- 
marians would not have called them SnrXa if ■ they were 
affricatae ;■ they could not possibly have classed them with the 
dirXa ypafifiara. 

This theory does not commend itself in anyway, and it cannot 
be consistently maintained, for when the combinations of aspi- 
rates come under consideration, it is found that it would be 
impossible to pronounce two affricatae consecutively {e.g. in 
p^^e?), and accordingly in these cases we are to suppose that the 
sound of the tenues was lost, and the <})9 or '^^0 were pronounced 
as two simple spirants. 

Before we pass on, we must look more closely into examples, 
already referred to, of two homogeneous aspirates being written 
instead of a tenuis and an aspirate, e.g. 'Bdxxot for BaA;;^o9. 

G. Meyer considers that writings like Baxxot show complete 
assimilation of the explosive k to the following x> ^^^ takes 
them as indicative of the last stage in the development of the 
Greek aspirates, that is, of their definite conversion into spirants, 
and that seems the only really plausible manner of explaining 
them. It might perhaps be argued that in these cases, as in 
those of the aspiration of a non-homogeneous tenuis to a follow- 
ing aspirate, the assimilation was merely one " of written letters 
and not of spoken sounds," but this seems. very improbable, 
especially as we further have instances of a single x instead of 
the double ones or tenuis and aspirate. Perhaps it would be 
as well here to enumerate the examples now extant of this 


;;^;^ for K')(^ we have from — 

Attica ex XaX;)(;tSo9 of 445 B.C. C.I.A. Slip. 27a 5, 17. 
'Idxx'p of third century B.C. (?) G.I.A. ii. 1592. 
Bdxxio'i G.I.A. ii. 1329. 

Boeotia Baxx^'>-^^"''i> c- 430 B.C. I.G.A. 157 ii. 14. 
Corcyra ^axx''^^^ G.I.G. 1850. 

Tenos Bap^;;;^o?, second century A.D. Ross. Inscr. lined. 
104, 3. 
also seven times Bap^j^^o? in Wescher and Foucart. 
And for single, x for kx — 

Ba;jj;i09, Ba;j^£aSa, Ba;!^t?, on a Delph. inscr. (v. Bull. Gorr. 
hell. v. 429). 

Of 66 for t6 we have from — 

Attica Ka.6dave ( = Kar{e)6av€) of fourth century B.C. 

G.I.A. 2719, 5. 
Mitylene Kd66eaav. G.I.G. 2169. 
Methymna KXeo06't? ( = KX6ot0/?). C.J.G^. 22115. 
Tanagra roe6lBa';. I.G.A. 157. 
twice on very old Corcyrean inscr. and coin "A.pa66o's, which 
Strabo vii. 325 and Pliny write "Apardo's. 
from Crete i66dvri &c., a-vve66a. 
Of ^0 for 7r<^ we have from — 

Mitylene on inscr. and coin 'Za<f><pw. G.I.G. 1211. 
Ithaca td<l>(f>ov, ib. 1927. 

"A<j)<j>€iv, ib. 3167 (c/. "Attach/, ib. 3278, 1), "Atf.^iov and 
"Acfxf}!], passim, for 'Att^t]. 
also single for 7r0 on a vase "Za^w, ib. 7759. 

As these instances of double and single aspirates are devia- 
tions from the correct spelling, they must be in all probability 
phonetic spellings. If this be allowed, it follows either tbat the 
Greeks sometimes spoke two consecutive homogeneous aspi- 



Chap. III. 

Indo-Germanic Origin of the 
Greek Aspirates. 

/O, 4>, X regularly correspond ^ change from hard into soft 
to : dh, hh, gh. aspirates. 

/their pronunciation 

^occasionally to : th, /where there is a surd Aspirate in 
ph, hh both Indie and Greek. 

.where the Indie has an Aspirate 
and the Greek a tenuis. 

possibly in isolated 
instances to spi- 
/ rants 

where there are double forms in 
Greek and it is uncertain 
/ which are the prior 

according to Curtius illustrates ease of 

Paul's caution. 

^Aspirated tenues (Brugmann, Curtius, &c.). 

.double change into surd spii'ants effected at once (Arendt). 
/probably spirants, or uffricatae, from the first (Briicke). 

close union of the two factors 
in Greek Aspirates. 

original Aspirate never, except 
in Asiatic dialects, superseded 
by tenuis. 

Chap. IV. 

Analogy of the Sanskrit 
Phonetic Laws. 

_/two consecutive syllables may not ^reduplication of Perfect in Sanskrit and Greek. 
begin with an Aspirate 

/Sanskrit compounds, 
/exceptions I 

I/Old Attic words. 

/a root may not begin and end 
with an Aspirate in San- 

the same rule generally 
observed in Greek. 

combinations of Aspi- 
rates unknown in 
/ Sanskrit 

ypvo Spir, /in Greek we have the groups <^d, x6 

/the Aspirate of a root retained before that of a termination. 

/ proAspir. | the first Aspirate in (j>e, xO, may not have been fully pronounced. 


rates (a combination which most authorities consider very 
difficult, or even impossible, to pronounce), or the aspirates 
had already become spirants, and the assimilation of the tenuis 
to the following spirant was a natural consequence, because it 
conduced to the ease of pronunciation. If this is the correct 
explanation, p^ and 6, if not ^, must have become spirants in 
Attica as well as in various dialects by the fifth century B.C. 

In the two forms Se8o/e;^0at from Samos {v. Cauer, Syll. 134, 
26) and eK^OefiaTu (Bull, de Corr. hell. vi. p. 262) we have a 
curious grouping of consonants which the supporters of the 
affrlcative theory take as a confirmation of the same. Others, 
however, seem with more reason to consider that we have here 
only cases of confused spelling, such as is also seen in ef?, 
Hvcr^dvTioi (Blass, p. 117). This latter explanation is probably 
also the best to apply to the writings /j.errjWaK'^oTe^ Alabanda 
(Bull, de Corr. hell. x. 302, 1. 39), avvSiaTrf^v\aKj(ev Mylasa 
(ib. v. 102), elaayeifOK'XpTa thrice (ib. xii. 84 f.) from 
Stratonikeia, and the converse writing fieTr)Xka^K6To<; from 
Pergamos (Inschr. v. Pergamos, Berlin, 1890, No. 225, 1. 16) ; or 
we might attribute them to uncertainty as to the exact form of 
the perfect termination. 



We will first review the relation of the Greek aspirates 
to the Indo-German, and point out briefly the etymological 
relations between them as demonstrated by Brugmann, 
G. Curtius, and others. 



We learn from them that % ^ ^ regularly represent Indo- 
German gh, hh, dh respectively, and that, originally at any 
rate, the Greek aspirates probably had the force of h, p, t with 
an appended breathing. Curtius further proceeds to show {v. 
ii. p. 17 £f.) that " though at first sight it seems that there has 
really been in this transition from gh &c. to hh &c. rather a 
strengthening than a weakening, and that it therefore seems to 
be an exception to the general tendency of languages " (which 
is one of the objections Roscher puts forward in his essay), yet 
that this change from soft into hard aspirates has now been 
shown to be one " which conduced decidedly to ease " of 
pronunciation, and does not therefore violate the general law 
of the development of languages. In confirmation of this he 
introduces and explains Arendt's theory (that the change con- 
sisted in a kind of assimilation of the first to the second 
element), and refers to Briicke and Sievers as having clearly 
shown that though the soft aspirates, were, and still are, pro- 
nounced in the Indian languages, yet they are sounds very 
difiicult to make. He further adduces the parallel to this 
hardening of the aspirates discovered by Ascoli in the language 
of the Gipsies. 

That the hard aspirates are easier to pronounce than the soft, ^ 
any one can readily test for himself by pronouncing " shep-herd " 
and " ab-hor " successively and without pausing between the 
mute and the " h!' 

It must be noted here that Curtius seems to attribute this 
change from sonant into surd aspirates primarily to a desire 
for ease of pronunciation, whereas we must not lose sight of 
the caution gvien by Paul {Princ. of Hist, of Lang. p. 47) that 
" the consideration of convenience in each production of sound 


affords in every case only a very subordinate and secondary 

If then % ^ ^ were, at any rate originally, a tenuis with a 
breathing added, they must have sounded much like the 
German tenues, only with the A-sound perhaps more dis- 
tinct; for it is well known that the Germans always aspirate 
the tenues preceding a vowel and saj', for instance, T-haube, and 
not Taube. 

Roscher is of a different opinion, and considers it to be far 
from probable that the sonant aspirates should in so many 
words have been changed into surd aspirates, if these latter 
sounded Hke the Mod. -Germ, tenues. 

Arendt says, "it is unnecessary to assume that aspirated 
tenues were really spoken for any length of time by the classic 
nations," and gives as his belief "that the double phonetic 
change, which transformed the sonant aspirates into surd 
spirants took place all at once." 

Briicke, while admitting the possibility oi y^^d having at 
one time been surd aspirates in Greek, advocates the proba- 
bility of their having been spirants, or at least aSricatae, 
from the very beginning. Brugmann's and Curtius' theory of 
the pronunciation of the letters is the one most generally 
received. If we now return to the question as to which Idg. 
sounds X'l'^ represent, we find they regularly represent the Idg. 
sonant aspirates gh, hh, and dh respectively. 

Sometimes, too, they represent an Idg. Jch, ph, and th, or 
a gh, and this especially in those words where the surd aspirate 
of the Greek is flanked by one in Indie, as in Koyx^o^ = Ind. 
9ankhas (v. Brugm. Grd. i. p. 406 K). 

In other words where, though Indie has a tenuis aspirate, 
Greek and other languages have a pure tenuis, it is a point of 

c 2 


discussion whether the Indie tenuis aspirate is here due to an 
original Indo-German tenuis aspirate, or whether it has been 
developed independently from a tenuis after the division of 
languages. If, according to the former view, all Indie tenues 
aspirates are to be traced back to Indo-Germ. originals, 
then we have to suppose that in Greek, as well as in other 
languages, the aspirate component of the Indo-German sound 
has often been lost. 

Again, if Bezzenberger and G. Meyer are correct in main- 
taining the priority of the aspirate in those Greek words where 
% and K, ^ arid v, and t appeat interchangeably after cr, then 
there are more instances than formerly admitted of the corre- 
spondence of Greek and Indie surd aspirates (v. Bzzb. Btr. 
vii. 63. ff.). 

In many cases, too, it is still doubtful whether the aspirate or 
tenuis is the prior, and also to what Indo-Germ. root they are 
to be referred. 

We may here draw attention to the fact that in no case 
where the form with an aspirate is acknowledged to be the 
prior, has this aspirate been permanently superseded by the 
by-form with a tenuis ; the two forms or the different words 
from the same root but spelt, some with aspirates, some with • 
tenues, have lived on side by side to this day. 

Now the fact of an original x <p never becoming a tenuis 
until quite modern times seems to argue a close union between 
the two component parts of a Greek aspirate ; they must have 
been pretty firmly welded into one, for otherwise in spite of the 
tendency to aspiration, we might expect to find a few words in 
which the original aspirate had become and Remained a tenuis 
seeing tliat they were so often interchanged. Even in modern 
Greek all the ancient aspirates are still intact ; in certain 


dialects, e.g. those of parts of Asia Minor, k, tt, t are spoTsen in 
place oi x^ ^> which seems to be a relic of the ancient Asiatic 
preference for the tenues, but in Greece proper x^ 6 are still 
spoken and writteu where the ancient Greek had a % i^ or ^, 
though in. the really popular language the second aspirate in 
the groups of two aspirates and also ^ or ;j^; after o- is now 
pronounced as tenuis. But that this modern tenuis has in 
these cases not arisen from and survived an original aspirate, 
while the form with the aspirate died away, is proved by the 
writings of the middle ages (when % ^ ^ were spirants) in which 
forms with tenues and spirants occur side by side, — oftener 
perhaps with the latter — thus ecfiOaae and ecjjTaae, da-Oeva and 
Spoaia-Tr}, a^VF^ (12 cent.) and o-koXij, idavixdaTrfv and 
eindcrOriKav, and so on. 

This sturdy persistence of ;;^ ^ ^ in the ancient, as well as in 
the modern language, appears to point to a complete amalgama- 
tion of the two component parts of the Greek aspirate. 

Curtius explains this by saying that " after the originally soft 
explosive had become hardened, not without the influence of 
the breathing, it would have been very surprising to find these 
hardened consonants again discarding this breathing." 

This sounds quite right in theory but does not seem to tally 
with facts as given by inscriptions, which yield us as many 
examples of tenues written for aspirates as they do of the 
opposite phenomenon. This causes us to ask, Why if these 
'' hardened aspirates " could frequently discard their breathing 
even in Attica (e.g. koXkovv, xirtov, cf. Meisterhans, § 38), 
without apparently any rhyme or reason for so doing, should it 
be surprising if in some cases the form with the tenuis had won 
the day and driven the aspirate entirely from the field ? 
Before leaving this part of the subject, it will be as well to 


discuss a few isolated words, in which the Greek aspirates 
irregularly correspond to Sanskrit spirants, and where the colla- 
teral Latin words have a spirant or breathing, so that, though 
the pure medial reappears in some of the kindred languages, 
we find the Greek x^6 flanked on either side by a spirant, and 
this induces the belief that they also may have been spirants, if 
only under certain conditions. 

Taking d first, we have the words x^mv, ydaiJ.aXo'i, with Skt. 
anm (sic Grassmann, p, 95) or Jcsham (Curt. p. 243), Lat. humi, 
Zd. zem,-, Slav, and Lith. zem-. Thus for this word we have 
spirants in all the kindred languages, and the Gk. p^^, Skt. x or 
ksh, Lat. h, Zd., Slav, and Lith. z all correspond, and are derived 
one and all from Idg. \/zhsenfb (v. Prellwitz, s.v.) or Idg. tjgzhom- 
(v. Brugm. Grd. i. p. 409), and either of these roots contains a 
spirant, so that for these Greek words the spirant is found not 
only in the sister-languages but also in the mother-language. 

Grassmann and Curtius explain it thus : the sfghama (from 
"which '^afiaC) first became " ghjam," and then by dentalism and 
assimilation " ghdham," and from this come xOov-, %^a/i-, and 
perhaps also the Sanskrit Jesham. 

They give the same explanation for %^69, Skt. hjas, Lat. heri, 
from " ghjes " as primary form : the development of this word 
Would thus be " ghjes — khjes — khthjes — khthes = ^^e'?. 

Yet in both these words the Greek d corresponds to an 
Indo-Germ. spirant, z orj, and in the Skt. and Lat. equivalents 
we also have spirants / or sh and h respectively, and for vOoiv 
the same spirant reappears in other languages, which makes it 
seem probable that, if only in this close connection with x,> & 
may have been a spirant. Similarly for x,> we often find it cor- 
responding to Skt. h, and Lat. h, as in %ta)i/, Skt. himam, Lat. 
hiems, exto, Skt. sahas, Lat. mho. , This Skt. and Lat. h generally 


represent, it is true, an original " gh," whose " g " reappears in 
Gothic, &c., thus e-xm is probably from an Idg. \/segho (v. Brugm. 
Grd. i. 422) — though Prellwitz derives it from tJsezho—\yvit the 
fact remains that in Skt. we have a guttural spirant A, and in 
Lat. a spirant or breathing " h," and, as there is some possibility 
of ■)(^ having been a spirant, these cases make such a supposition 

In the word ^payf^fiav for Skt. brahma (v. Strabo, xv. 1, 59, 
Diod. xvii. 102, &c.) the medial ■)(^ represents the Skt. spirant 
" h," which could perhaps justify the conclusion that by these 
historians' time, that is, by the end of the pre-Christian era, ■x^ 
had already become a spirant. For if ■x^ had still been a tenuis- 
aspirate, Ich, the Skt. spirant " h " would have probably been 
altogether omitted in the Greek rendering of the word, as 
^pafiaP6<; would have sounded more similar to the Sanskrit 
original than j3pakhfidve^. 



Amongst several reasons which Curtius instances as apparent 
proof that % ^ ^ were real aspirates is " the movable nature of 
the breathing which (a) is easily separated from the explosive 
element, and leaves the explosive element behind, e.g. 7re(j>vKa 
for <j}e<j)VKa, but (6) just as easily unites with another explosive, 
and though its position varies, does not do away with the feel- 
ing that forms like dpi-ilrco and Tpe<f)(o belong to each other." 

Of these he proceeds to say : " I doubt whether such pheno- 
mena occur in any language with recognised spirants, whereas 


the first two " (those we have quoted) " have exact analogies in 

We do not think that the analogy in question can be con- 
sidered at all in the value of proof, but at the same time it seems 
worth statement and investigation, merely in so far as it is an 
argument from analogy and one that has been supported by such 
high authority. With this premiss we will now compare the 
Sanskrit and Greek phonetic laws. 

An important phonetic law in Sanskrit is, that two consecutive 
syllables of a word should not begin with an aspirate, and, to 
avoid such an occurrence, in the reduplication of verbs for in- 
stance, the breathing of the first aspirate is dropped and leaves 
the explosive element behind, e.g. da-dhami from Jdha. 

This fact, which speaks strongly for the real aspiratic nature 
of the letters in question, is also generally observed in Greek, as 
for instance in the reduplication of. verbs, e.g. iretfiiXriKa, not 
<^e-^i\r]Ka, from (piXeo), and ridrj/ii, from Jde. 

The same principle of the movable nature of the breathing is 
evident in the law, known as Grassmann's, which is common to 
Sanskrit and Greek, though unknown to Indo-German, and which 
forbids a root to begin and end with an aspirate. Thus Idg.^O/ieudh 
becomes Skt. ijhudh, Gk. i^irvO ; compare also Skt. dhalc- from « 
njdagh, Gk. Tpt,')(p'i from Qpi^, erd^rjv from Odinm, &c. 

This process of dissimilation, which makes one aspirate 
reappear when the other disappears, does not, as Curtius says, 
do away with the feeling that forms like dpeyfra and Tpijxo 
belong to one another, and certainly argues powerfully for the 
similarity of pronunciation, at least originally, o{ ■^^cjid and the 
Skt. aspirates. For ii ■^cfid had from the earliest days of the 
Greek language been real spirants, we might have expected to 
find in the reduplication of verbs two consecutive spirants, 


as we do in Latin fefelli from fallo, and Gothic hvai-hvdp from 

Of course it may be objected that the Greek ideas of euphony 
and kakophony may have differed widely from those of the 
Latins and Goths, and that they may have found the sound of 
two consecutive syllables beginning with spirants very disagree- 
able, and consequently replaced the first spirant by a tenuis. 
However this may be, the fact that in two such important 
points Greek corresponds with the Sanskrit usage, apparently 
indicates that at some time or other the Greek aspirates were of 
the same nature as the Indian, but it is not sufficient to prove 
that they still were so in the classical period of Greek literature, 
as Curtius says. For supposing that x'^^ were already spirants 
at that time, even so these forms of verbs, nouns, &c., had long 
before that time become stereotyped in the language, and would 
not have been aifected then by any change which was taking 
place in the pronunciation of single letters, any more than they 
have been since throughout the many centuries during which 
X4>0 have been undoubted spirants — Tpe<^<o and dpiyjra) are 
still said. 

This rule we have under consideration, namely that aspirates 
should not stand at the beginning of two consecutive syllables, 
is not violated in Sanskrit, except in compounds, e.g. abhi-bhutis, 
ahi-han, and at first sight it appears to be strictly observed in 
Greek also, but on closer inspection one finds it is by no means 
inviolably observed. In old Attic, Meisterhans tells us (c/. 
Gram. d. att. Inschft. § 37) this law " does not seem to have been 
firmly established," as on inscriptions of the fifth and fourth 
century B.C. forms are found, such as e^ei (perhaps by analogy 
with e^co), C.I.A. iv. 373 ; ivOavdol, ib. iv. b. 27, <pap6evo<; and 
. YoX.%09, ib. iv. b. 373, &c., and sometimes even three consecutive 


syllables commence with an aspirate, e.g. 6v(f)ai6iBr]<;. And apart 
from errors on inscriptions, we have regular forms as eddcpOriv, 
Te0d<jidai (by the side of irdi^rjv), i(l>dv6r]v, (/)a^i, and others. 

These exceptions, which are fairly numerous, seem to indicate 
that the Greek aspirates were already diverging somewhat from 
the Indian. 

Another strict rule in Sanskrit is, that two aspirates should 
never occur together, but that an aspirate may only stand before 
a vowel, semivowel, or nasal, and an exception to this rule 
appears perhaps onjy in the word Vifhthala, which is probably 
not of Aryan origin. In Greek, on the contrary, this rule is not 
observed, as there are numerous instances of two aspirates coming 
together, not only medially, but even initially, e.g. i'^dpo';, ')(6(!t)v, 
^Gelpa — such combinations are quite unknown to Sanskrit. 

The same dislike to the union of two aspirates in Sanskrit is 
shown by the rule that " when the final sonant aspirate of a root 
is followed by a t- or th- of an ending, the combination is made 
sonant, and the aspirate of the final is transferred to the initial of 
the ending, e.g. tjrundh+thas becomes not rwwdhdhas, but" and njladh+ta becomes 6addha {cf. Whitney, /S^if. Gr. 
p. 50). 

In Greek we have an exactly opposite phenomenon, which is 
difficult to explain on the assumption that the Greek aspirates 
had the same phonetic value as the Sanskrit — for here the rule 
is, that (1) " when a final surd aspirate of a root is followed by a 
th of an ending, both the aspirate of the final and that of 
the initial of the ending are retained," e.^'. <yypa<j) + dr]vai,= 
ypaipBrjvai ; and (2) that " when a final unaspirated surd of a 
root is followed by a th of an ending, the unaspirated surd 
of the final is aspirated, even though it belong to a different 
class of surds,'' e.g. y/SeK + Ofjvai becomes SexOfjvai, whereas 


if had the phonetic value of i + h we should have expected 
heKOffvai to be written. 

Thus also, contrary again to Sanskrit usage, when a final surd 
aspirate of a root is followed by an unaspirated surd of an 
ending, this aspirate of the final is not transferred to the 
initial of the ending, but is entirely dropped, e.g. ,y/ypa<j) + T09 
becomes not 'ypaTr6o<;,hvit 7pa7rTos, whereas in Sanskrit Jbadh + 
ta becomes baddha, not iatta or badda. 

This regular aspiration of the explosives before 6 is cer- 
tainly the point which is most difficult to understand, if we 
accept the aspiratic theory of x^O. Curtius comes to the con- 
clusion that " no definite argument is to be drawn from grouped 
aspirates," and that Von der Miihll's notion that the assimila- 
tion was in these cases only one of written letters, and not of 
spoken sounds, is deserving of much consideration. 

But this grouping of the aspirates is such a distinctive feature 
in the Greek language, that this summary explanation of it 
seems hardly conclusive and satisfactory. In dealing with dead 
languages, writing must, if not wholly, yet largely "be our 
evidence for the original sound," and the constant, regular 
recurrence of a peculiar orthography is surely of immense 
importance in determining the sound of the individual letters. 
This axiom is one generally accepted, and the arguments for 
discarding it in this one particular case do not appear weighty 
enough to justify this being done. 

From this assimilation of a tenuis to a following aspira,te, 
which except in four or five instances is uniformly observed, it 
may be justly inferred, we think, that the tenuis did not retain 
its original sound in this position, but that some distinctly 
audible change must have taken place in its pronunciation. 
For if the ^ or ;\; in i^6, ^S still continued to be spoken as a tenuis, 


or was so slightly aspirated that the aspiration was practically 
inaudible, how is it that the inscriptions, at least the private ones, 
do not afford numerous instances of exceptions to this artificial 
orthography yd, (j>d ? The four or five exceptions, such as 
airdLTOv, " found on archaic and later monuments," seem 
hardly worth mentioning, as Blass says (vide p. 103), when com- 
pared with the countless examples still extant of the omission 
and misplacing of the spiritus asper, due to its weak sound ; so, 
too, if (pd, yd were practically pronounced as ivQ, k9, we ought to 
have, not only half-a-dozen, but many dozen instances of such 
miswritings. It appears somewhat incredible that every stone- 
mason should have a sufficiently critical eye to discern that it 
would look incorrect, or that he should remember his schooling 
so well as never to forget that it would be inconsistent with the 
law of assimilation to write a tenuis before an aspirate, and 
therefore always carefully wrote two aspirates, even though he 
pronounced the first of the two exactly, or very nearly, as a 

In short, that their spelling should have been so consistently 
inconsistent with their speech, is hard to believe, and if other 
evidence tends to prove that % <^ ^ were real aspirates, it seems 
more logical to decide that in these groupings also two aspirates . 
were spoken as well as written, even though it may be quite 
impossible for an Englishman to pronounce fhth or khth. That 
it is not an impossible combination to some nations we learn 
from Sievers {Lautphysiologie, p. 96, note), who says that " Die 
deutschen Mundarten .... die doppelte Aspiration vermeiden. 
Ich bemerke dass aber anderwarts z. B. im Armenischen, diese 
Abneigung nicht besteht und man wirklich zwei nicht homorgane 
Aspiraten neben einander spricht." 

We have thus no reason for asserting that two consecutive 






Chap. V. 
Testimony of tho Grammarians, 



/vahie of their evidence 

Chap. VI. 

Testimony of Classical 


I/depends a good deal on its date. 

/ttc/xui'a and yjfxitfioiva 

/oi various terms and 

^inaccuracy and variety of definitions 

paralleled by 
/-that of modern grammarians. 

/-Aristophanes' I 

of definitions of in- 
^ dividual letters. 

/j)TO Aspir. 
analogy of German and Hungarian 

yirviv/xa as used by Dion. 
Halic. and Schol. 
on D. Thrax 

different descriptions in- 
applicable to real 
/ aspirates 

X®. D. H's. and Schol, 's on D. 
Thr. defs. can apply to 
the spii'antic theory 

/to be rendered "breath" not "breath- 
ing " = spiritus asper. 

yrrpotrdi'jKri rov Trj/ci'yuaTos, not to be taken 


cf. Walker and Sweet on Eng. th. 

■pro Spir. 

Spirants of one nation constantly 
pronounced as homogeneous tenues 
'' by another. 

'Plato's TTvev/jiaTw&rj | not definite enough to allow of any conclusion. 

, *. D. H's. and Arist. Quint.'s 
defs. applicable to a bi-labial 

/X. D. H. and A. Q. vagn 

Schol. on Dion, Thrax seems most 

Schol. on Dion. Thr.<ix describes a 
spirantic ^. 


Chap. VII. 
Early Orthography. 

the writing IIH, KH /why abandoned, if it denoted the phonetic value 
l/we do not find TH. 

/the writing ®H 

/may denote aspiration. 

I/paralleled by insertion of A after other letters. 

/pro Aspir. ko", -tto- another substitute. 

yX"'' "^o- = ^> V \/pro Spir. /the Naxian hs as substitute for ^o-- 

1/some raciis may have heard a softer spirantic element in the first factor 

cf. German chs. 


aspirates would have been a combination impossible for the 
Greeks to pronounce, and the supposition that they did speak 
two aspirates is more satisfactory than Von der MiihU's 

And yet if they were both pronounced as aspirates, should 
■we not have more instances of tenues being miswritten for one or 
both of them, seeing that the interchange of tenuis and aspirate 
is so very frequent in the case of ^ngle letters ? 



We now enter upon our especial investigation of the internal 
evidence of Greek itself, and will begin it by estimating 
the testimony of the Grammarians as to the nature of x4'^- 
Unfortunately they are not very accurate in their definitions 
and descriptions of the different letters; some of their expla- 
nations can easily be interpreted in different ways to suit 
different views, and others are obscure and therefore but of 
little help to us. The only way of getting an insight into their 
meaning is, we take it, not to quote, or read, a few isolated 
lines which bear upon the particular letters under discussion, 
but to read all they say about the letters, which is never very 
much, and try to catch their spirit and look at the letters from 
their point of view and not our own. We should further take 
into account the times in which these different grammarians 
and other writers who mention the subject lived, as the worth 
of their evidence for the pronunciation of the classical period 


depends a good deal on this — and it has not heen sufficiently 
taken into account as yet. 

So little is known of the lives of the different authorities 
that their dates can only he very roughly fixed, hut if -we do 
this, we shall find the following to be their chronological 
order : — 

r Dionysius Thrax . . . fl. B.C. 80. 
\ Dionysius Halic. . . . „ B.C. 30. 
' Aristides Quintil. . . . „ a.d. 100 ? L. and S. 

„ 3rd cent. Blass. 
Sextus Empiricus . . . „ A.D. 220. 
Diogenes Laertius . . . „ A.D. 280 ? 

Only two therefore lived in the pre-Christian era, and that long 
after the classical period of Greek literature, whilst the others 
all wrote at a time when % </> ^ were almost undoubtedly 
spirants. Blass does not seem to take this fact into con- 
sideration at all, as he says Aristides Quint, lived in the third 
century, then quotes his words, and immediately adds " accord- 
''■'^9^'!/ X ^ ^ were instantaneous and explosive " ; " accordingly " 
here meaning " according to Dionys. Hal. and Aristides Quintil.," 
and yet these two were separated by an interval of at least one- 
and-a-half centuries, so that their evidence ought not to be » 
taken as of equal value, and if Aristides Quint, did not live 
till the third century A.D., % ^ ^ had by his time almost certainly 
become spirants. 

The scholiast on Dion. Thrax, whose remarks can profitably 
be compared with those of Dion. Halic, is of very uncertain 
date, probably between 500 and 700 A.D. ; at any rate of a time 
when x4'^ vrere spirants. 

We will now proceed to sift their evidence and see what we 
gain from it. 


Dion. Thrax and Dion, Halic. both reckon % ^ ^ among the 
mutes or a(f>a)va, and the latter defines mutes aright as being 
those oaa ovre Ta<; reXeta? oiire ra? ^/itreXeia? <f)(ova<; e'^ei 
Kad' kavrh, fied' erepcov Be i/ctjxoveiTai, and the rjfii^ava are 
those oaa fjtera /lev rmv ifxovrjivTfov Kpelrrov iK(f)epeTai, Kad' 
eavTo, Be -^elpov re koI ovk avToreXS)';, and to this class, he 
says, belong among others X, yit and v. These liquids are so 
called because in them there is no complete stoppage of the 
breath, whereas in y(^ ^ 6, if real aspirates, there would be 
complete stoppage at first. However, that some uncertainty 
existed as to the class to which % ^ should belong, is shown 
by the fact that the Stoics (according to Diog. Laert.) did 
reckon them among the '^fiicjxova. But^ even ii ■^ <f) were 
spirants, they would very likely from force of habit and accord- 
ing to some old-established division of the letters, made perhaps 
when the old writing IIH and KH still existed, have been 
classed with the mutes. That the term a^wva was certainly 
used in a very loose manner is shown by the fact that Dion. 
Thrax, the only other pre-Christian authority, gives a very 
different definition of the word from that given by Dion. Halic, 
and one that is practically of no use for phonetics, for he includes 
all the letters except the seven vowels under the name a-vp,^a>va 
and then divides these into rip,i<ficova and d<j)<ova according 
to their being more or less euphonic, thus : a,(f>co va Be Xiyerai 
OTi fiaXXov T&v aXXcov icrrl KaKoipava, wairep a<f>o)vov Xeyo/iev 
rpaypBov tov kuko^wvov, and again, '^ p, i (fi to v a Be Xeyerai on 
trapoa-ov ^ttov twv ^minjevrcov evifxova KadecrrrjKev iv Toi<; 
p,vyfiol<; Kol cn.yp.ol's, about which the Scholiast very justly 
remarks : koI evravda a<f)cova Xeyerai w? KaKo^eova, kuI ovk w? 
reXeo)? (fjeovrj'; ecrreprj/ieva. 

And this is our oldest authority, who is often adduced as a 


witness for the aspirate pronunciation of ;j^ ^ because he places 
them amongst the a<f)Q)va or mutes. He certainly reckons 
them among the dcpwpa, but whether his atjxapa mean what 
our " mutes " mean, must remain an open question as long as 
our knowledge of what he meant by a(f>cova is limited to his 
definition that I have quoted — in other words, as long as we do 
not know what sounds he would have called ev(j>cova and 
which KaKOJxova. 

The evidence of Dion. Thrax is but of very little value to us, 
therefore, as no argument can be based on premises which are 
indefinite and not accurately stated. 

Aristides Quint, does not define d<p(ova at all. 

Another passage which shows the wide sense allowed to the 
term d<f>a3va is Plato, Theaet. 203 B, where he says : to a-ty/ia 
rS)v dtfxovcov earl, i}r6(j)og n<; fiovov. . . . ' that is, he reckons s 
among the dcfxava and calls it " a mere noise " and goes on to 
say "h and most other letters are neither vowel-sounds nor 
noises. . . . The most distinct, which are the' seven-vowels, have 
a sound only." Brilcke says this proves that Plato understood 
under dcjieova those letters " in which the sound of the voice 
was wanting," i.e. our consonants. Further, it must be noted 
that though here he places cr among the d<p(ova, he evidently ^ 
knew of the subdivision rj/jLL^cova as we see by referring to 
Crat. 424 C, where he divides the letters into three classes, 
(1) (fxovijevTa, (2) a^cova koi dejjdoyya, (3) (fxov^evra fiev ov, oil 
fievToi, ye. d<^6oyya, 1 ( = ^fii<j)a)va). 

But really it is perfectly needless to argue on this point, as 
the science of phonetics had certainly not by the end of the 
pre-Christian era attained that accuracy which it has now, and 
subdivisions of letters into sibilants, fricatives, &c., and the 
distinction between aspirates and spirants had very likely ne.ver 


been heard, or thought of, and of the five or six Greek writers 
who mention any classification only the two, Dion. Hal. and 
Dion. Thrax, can lay any claim to the name " Grammarian " — 
the others may have known no more of " phonology " than any 
ordinary well-educated man or woman, — so that, even ii ■)(^^d 
were spirants in their time and yet they classified them under 
mutes, it does not appear very astounding. For if a man in 
1892 can classify the English spirants /, v, th under mutes 
although they have alwaj's been spirants, and never aspirates, 
in his language, then it is quite comprehensible that a man 
of B.C. 30, who had by no means the same opportunities of 
learning the correct phonological name to be applied to the 
individual letters, should also classify under mutes letters which 
in his own time were spirants, but had in a former period of the 
language been aspirates. 

And in order to see that English spirants are still classified by 
some under mutes, we need only turn to the twenty-fifth edition 
of Adams' English Language, pp. 59-61, where we shall find 
that/, V, th are classed with^, h, i under the big heading " Mutes," 
and are called aspirated mutes or aspirates, and this in a book 
published in 1892. . 

The Stoics, however, apparently considered ;)^ ^ ^ to be r)fi,i^(ova 
(Sext. Emp. is uncertain whether to call thera'ai/xai/a or y/nicjicova), 
and to show that, even if spirants, they might well have been so 
called, we can refer to Murray's English Grammar, 1853, where 
/, V, th are called " semivowels " (vol. i. p. 35). 

The mistake of calling English th, v, f "mutes" and 
" aspirates " does not occur in all English grammars of course ; 
Mason and Morris, for instance, call them " spirants '' and 
" continuous consonants " respectively — yet the fact that any 
grammarians should still call them " mutes " and " aspirates " 



shows the direful confusion which still exists as to the true use 
of these terms. 

Supposing then for the sake of the argument that X'P^ '^^^^ 
spirants in the time of Dion. Hal. and the Stoics, we have aa exact 
parallels to their calling them a^wva and ■^fiicjiwva respectively, 
the case of two English grammarians of this century, who call 
the English spirants " mutes " and " semivowels " respectively. 

Of course this evidence by no means proves, or is intended 
to prove that y(^<j)d in Dion. Hal.'s time were spirants, but it is 
only adduced to prove that it is utterly futile to try and base any 
argument as to the specific nature oi ■x^tj) on the fact of their 
being called by two grammarians a<^aivd and by other writers 
not grammarians, either d(f>(ova or ■^fio(j>o)va. 

We see then that if the word " mutes " is still often used in a 
loose and improper manner, the term d(j>a)va may well have been 
similarly used in an age when only very rough and large classi- 
fications of the letters had been made. 

We will now pass on to the examination of Dion. Halic.'s 
other remarks, and will first take the passages in which the 
word TTvevfia occurs, to see what he meant by it. 

After speaking of the vowels Dion. Halic. says : — 

(a) p. 75 rovTwv Brj (i.e. of the vowels) <f>eovrjv rjSiarrij/ 
diroTeXei ra /jLaKpd...0Ti iroXvv ^■^elrai "^ovov «ai tou irvev- 
fia T o <; ov KaraKOlTTei rbv tovov. 

(b) then on p. 83, rpia (lev eKcfxavelTai utto twv veiXewv 
cLKpmv, TO TT (j) 13, orav, tou arofiaTO'i •jne<r6ivT0<!, t o 
IT p o ^ aXX 6 fi 6 V o V € K Trj <! ap T r) p i a<; ir v ev fia Xvay 
Tov SecTfiov aiiTov. 

(c) p. 84 : fiia /x,ev avrr) crv^vyia rpowv jpafifiaTcov a(\>wvav 
o/ioim <7y(r]/xaTi Xeyofievmv, •yjnXorriTi 8e Kal BacrvTrjTi (tov 
TTvev/iaTOf) Bia(j)ep6vT03v. 


{d) ih. Tpla he aXKa Xeyerai, t^? yXaiaarjis dxpo) to3 
GTOfJiaTL "TrpotrepeiBofievi^^ Kara tov? ixerewporepov; 6S6vTa<;, 
kireid' V IT o r o V "Trvevfiaro^ v'iroppa'in^o/j,evri<;, koX ttjv 
8ie^f)Bov a V T m irepl tov<; oBovra^ d'7roStSovar]<;, to t koL to 
6 KaX TO B 

(c) ib. Tpia he to. Xotira tcov a^mvcav XeyeTai fj,ev, ttj? 
yXcoTTTji; avKTTa/jievrj'i kuto. tov ovpavov e'^jv^ Tfj<; ^dpvjyo<i, 
Kol Tr}<} apT7]pia<; virri'^ovari'; Tm irv ev p,aT i, to k ')( y. ovhevl 
ravTa hia<j)ipovTa tcS ay^ijfiaTi aXXijXtov, TrXiju oti to /j,h> k 
■\}nXa)<; \eyeTat, to he ^ haaeta^, to he y /MeTpieov koI p,eTa^v 


(f) p. 85, he adds : xpanaTa fiev ovv iaTlv oa-a t £ ir v e v- 
p, a T I TToWoS Xeyerat" hevTepa he, oaa p, e cr m (tt v e v p a t !,)• 
KUKiQ) he, oa-a i^t\o3 {tt v ev p ut i) TavTa pev yap ttjv eain&v 
hvvapLLV eyei povrjv Ta he haaea kuI triv tov it v e v p ut o<; 
•jr p o a 6 rj KT) V, ci)? efyYV? tov TeXeioTUTa elvai. 

If these passages are read through one after the other with 
unbiased mind, the best and only really possible translation of 
TTvevpa will, we think, be admitted to be " breath " or " stream 
of air," and as, when the whole description of the letters is read 
connectedly, it is quite clear that the writer uses the word -jrvevpa 
throughout in one and the same sense, it should also be so trans- 
lated in every case; In extracts (a), (6), (d) and (e) it is self- 
evident that it means "breath" or "stream of air," then apply- 
ing this same translation to extnact (jc) we get that " the first 
set of three mutes which are pronounced with the same vocal 
organ, are distiuguished from each other by the weakness or 
strength of the stream of air [or breath] " used in pronouncing 
them and NOT that they are distinguished by the weakness or 
strength of the " breathing," with " breathing " taken to mean 
the " spiritus lenis " or " fortis" respectively; and siiriilarly extract 

D 2 


(/) must be translated :—" The strongest are those, which are 
pronounced with a strong stream of air [or much "breath"], the 
second, those pronounced with a moderate one and the weakest 
those pronounced with a scanty amount of Ireath [weak one] ; 
for these last letters have only their own force, whereas the Sao-ea 
have the help of the stream of air [or " the addition of the 
breath "] and thus come near to being the most perfect letters." 
And when he says they are near being the most perfect 
letters he is evidently comparing them with the vowels, and the 
reason he has given for their being such is because " they 
do not check the current of the breath [or "interrupt the 
sound of the voice "] or stream of air." Here certainly no one 
would wish to translate " tov TrvevfiaTc; " as " a breathing i.e. 
an h," for how could it be said of the long vowels that " they 
have the pleasantest sound because they do not check, or break 
off, the sound of the ' A. '" ? Then why in the case of the letters 
Xi^^, which Dion, is evidently mentally comparing with the 
vowels and thinking they are very nearly as perfect as the 
vowels and only a little inferior because they have TrpoadijKrjv 
TOV TTvev/jLaTOi whereas the vowels are altogether Trvevfia, — why 
in this obvious comparison should the crucial word on which the 
whole comparison is based be differently translated in the twq, 
branches of the comparison ? Why in the second clause should 
irvevfia mean a " breathing i.e. an h '' when it cannot possibly 
bear the same meaning in the first clause ? And, as a side- 
question, which are generally considered to be more similar 
in nature to vowels, aspirated mutes or spirants ? Evidently 
spirants, if they " according to the ancient nomenclature ought 
to be called rjfil^wva" (cf. Blass, p. 100). 

From this consistent and impartial translation we get a 
description of the d^cova which, as Blass also finds, certainly 


suits the modern pronunciation of tt /3 &c., even though it may 
not prove it, but neither does it prove that % ^ were real 

Dion. Thrax does not himself use the word irvevfjia in his 
short account of the letters, but his scholiast does and, as he 
wrote when ,8 and </> were spirants, it will be instructive to 
compare his use of Trvevfia with that of Dion. Halic. 

Dion. Halic. Scholiast. 

KpaTiara. . .ocra too tt v e v- haaea he...ra 7roX\c3 

fiaT I TToWaj Xer^eraf ir vev fiar i eK^wvov- 

Sevrepa Se, baa fie (7 m- fieva- fieaa Se, t^ fi-^ t e 

KaKico Be ocra ^jr iX £. ir oWa fii] t e oXi y a> 

(—fiea-a)) yjriXa Si XeyeTai, 
TO, 6X I y a) TTvevfiart. 
and again 

ovSevi T a V T a («%7) /car' oiSev yap hiac^epet to 

8 lacj) e po V T a t£ ^XV- "" "^^^ ^> ^' A'V o''"' fie t a 

fiari aXXrjX(ov,TrXr]v tt oXX o v tt v ev fia t o <; 

o r I TO K yjriXa)<: Xiy e- eK(j)(oveiTai. 
T at, T b Be x B aa e (o <s 
( = iroXXS TTvevfiari). 

These descriptions tally in a wonderful degree if we are to 
believe that one of these writers still pronounced % ^ ^ as 
aspirates and the other pronounced them as spirants : they 
seem rather to point to a pronunciation common to both, 
always supposing that in both cases we translate Trvevfia by 
" stream of air " or " breath." They both alike lay stress on 
the fact that the i/rtXa and Baaea only differ from each other 
in that the latter are pronounced iroXXm irvevfiaTi. Further, 
here should be noticed, what seems to have been quite left out 


of sight, that these old grammarians counted letters as " in no 
wise differing " as long as they were pronounced " with the 
same organ of speech " — eK^wvovfieva Kara tov avrov rpoirov 
TMv (jxovriTiKaiv opydveov. Thus, because tt was an d^covov 
and pronounced with the lips, /3 and ^ being also labials would 
come under the same heading as tt, i.e. under the first set of 
d^cova, and this regardless of whether they were mutes or 
semi-vowels (■^fj.i<pQ}va.). Similarly t B 6 and « 7 % being 
placed in two distinct sets as dtjxova does not show that 
they were all real mutes, but only that for t S ^ the tongue 
was in one and the same position, and in a different one 
for all the three « 7 pj;. That this was so is proved by the 
fact that even the scholiast of Dion. Thrax says that tt and 
<f> are only distinguished by the ttoXXw irvevixaTb of the latter, 
otherwise they are the same, because they are eK<j}ci)vovfj,eva 
Kara tov avrov rpoirov rcov (pavrjrtK&v opydvav ; he nowhere 
says that (p was not an d<p(ovov but an rjfii^covov, although 
it is allowed by all that in his time was certainly a spirant 
or Tjij,i(j)(ovov. Aristid. Quintil. also begins with the three 
that Sia r&v y^^eiXewv f]-)(elrai jjlovov, and goes on to the 
dentals and gutturals. 

As the grammarians divide these letters into three sets 
according to the organ with which pronounced and make no 
further distinction between tt and (p, k aud p^;, r and 6 beyond that 
of the breath, Blass concludes that % <^ ^ were, like Kirr, ex- 
plosives because, as he says, "in the modern Greek pronun- 
ciation no one could ever maintain these letters to be mutes." 
This appears to be a somewhat too hasty conclusion, for the 
English /, V, th are quite as genuine spirants as the modern 
Greek x.4> ^> ^^^ 7®*. ^ we before pointed out, they are in 
many grammars of the present day classified as mutes and, just 


as the old Greek grammarians divided their nine mutes into 
three classes according to the different organ of speech em- 
ployed, so nowadays we still find grainmarians dividing the nine 
mutes jp, b, f, and v ; t, d, th ; k, g, Scotch ch ; into three classes 
" according to the part of the mouth chiefly used in pronouncing 
them," i.e. into labials, dentals, and gutturals, and no further 
distinction is drawn between them except by saying that they 
are " sharp or flat, aspirated or unaspirated." Vide Summary 
of English Grammar, compiled for Notting Hill High School, 
third edition, Rivington, 1890, p. 5. Could we have a more 
exact parallel to the procedure of the Greek grammarians ? 

Thus the fact of the grammarians including % ^ ^ under the 
mutes does not exclude the possibility of these letters having 
already been spirants at that time, or, in other words, no 
argument can be justly based on their classification of mutes. 

To go back to the word irvevfia. This, in the last extract 
from Dion. Halic. where he speaks of rrjv irpoaOiqKrjv tov 
trvevfiaro<;, ttoXXoG, ixia(p and ■yfriXw Trvev/J-aTi, is often trans- 
lated by " breathing " in the technical sense, and Trpoad^JKi] is 
translated " addition " in its baldest sense, and thus TrpoaO'^Kr/ 
TOV irvevfiaro<i is taken to mean " addition of a breathing, i.e. 
an h " (cf. Blass, p. 99) ; if this is the correct translation, and 
the two sounds of the tenuis and spiritus asper were distinctly 
heard and had not coalesced into one, then it is perhaps strange 
that not a single writer mentions the fact that they have a 
double soiind, as Dion. Halic. does oi ^ ^ yjr. He particularly says 
that these are ypdfifiara SiifKa a fiiKrhv Xafi^dvei tov yp-o^ov, 
then why does he omit to mention that ■)(^<f>d axe also hmXa ? 

Then again, if -n-vevfia means ' breathing,' w k t, are pro- 
nounced with a ■y^iXm nTvevfiaTi = spiritus lenis ; cj) ■^ with 
TToXXw TTpevfiaTi (or Baa-eat;) = with spiritus asper ; and /S 7 S 


with fieam TTj/eu/iart = with half the spiritus asper, or, ex- 
pressed differently, — 

IT = Engl. p. 

/3 = „ b + I h. 

</) = „ p + h. 
Even though there is a distinct difference in sound between a 
pure tenuis or media and an aspirated ditto, it is difficult to con- 
ceive of a half-aspirated letter, and would a nation have been able 
to go on for centuries pronouncing half-aspirated mediae without 
converting them into real aspirates or pure mediae ? Even if it 
did achieve such a miracle, one would imagine that p and p + h 
would have sounded most alike of the three and that /3 would 
not have been pushed in between the two in a classification. 

But, in truth, it does seem as if, in connection with the 
" mediae " especially, -nvevfiaTi cannot bear the forced meaning 
of " breathing i.e. an h." If it must bear this meaning, then 
/8 7 S were " half-aspirated " sounds, that is half-aspirated sonant 
explosives, and if so, it appears incredible that they should 
have existed as such any length of time in Greece. For, as we 
learn from Curtius and others, the Indo-Germ. sonant aspirates 
were certainly difficult to pronounce, and consequently not 
preserved in Greek and other languages of the Indo-Germanic 
stock, but converted into surd aspirates or pure mediae. Now, if 
this was the fate of the fully-aspirated sonant explosives, is 
it reasonable to suppose that a language which rejected them 
should for centuries have retained, if not actually introduced, 
half-aspirated sonant explosives ? We think not. We cannot 
consequently reconcile ourselves to the idea that in these extracts 
•7Tvev/j,a should mean " breathing," Trveufj^ari iroXXm = " the 
strong breathing," &c. It seems a useless perversion of the 
meaning of the word, which after all does not give a satis- 


factory proof of the theory it is required to demonstrate, 
whereas, if we give it its ordinary and natural meaning, then, 
though it certainly seems to adapt itself better to the spirantic 
theory, it still remains vague and indefinite and can be used for 
either. It is also difficult to understand how Dion. Halic. 
could speak of the Saaea as being nearest to the most perfect 
letters, the vowels, if they were pronounced as distinct tenues 
with a breathing, or as KpaTiara, strongest, if they were a 
weakened form of the tenues. 

Aristid. Quiutil. (in whose time it must be remembered % (/> ^ 
were probably spirants) says : ja 8' evSodev ex (j>dpv'y'yo<; 
d)v6/j,a(7Tai haaea kclv iari, ~Kiav rpw^ea. Aspirated tenues do 
not answer to this description either very well, as they cannot be 
called \iav rpa'^ia, whereas this epithet applies very well to 
the spirants ; every one will doubtless admit that in a%09, for 
example, the sound is a much rougher one if it is pronounced 
" Sichos " with Germ, ch than when pronounced ak-hos ; simi- 
larly in Adrjvai = Athrjvai. with Engl, th, or = At-h^i^at, the 
spirantic ^ is a much rougher sound. The breath, too, is not 
brought energetically from the larynx in pronouncing the 
aspirated tenues, but simply from the front of the mouth, 
whereas for the spirants it is most decidedly brought ener- 
getically from the larynx and the rough sound is produced. 

As regards the individual letters 6 cf) and ^, it is generally 
taken that the description of Dion. Halic. proves them to have 
been aspirates in his time, whereas that of the Scholiast of 
Dion. Thrax shows that in his time they were spirants. We 
will take them singly and examine what is said about each. 
0. From extract (d) from Dion. Halic. (p. 33) we see that for 
all three letters alike rSd he says " the tongue is pressed against 
the upper teeth," whereas the Scholiast says that for 0, t^s 


yKcaa-arj'; air o'^(o pov a rf <; t & v 6 B 6 v r (o v Koi Trape'X^ovarji; 
e^oBov To3 TToXXeS TrvevfiaTi- — and hence it is argued by Prof. 
Blass (p. 99) that 6 according to Dion. Halic. is an explosive, 
not a fricative, " because the tongue is not pressed against the 
teeth in making the th-sound, but only brought near," just 
as the Scholiast says that for 6 the tongue is no longer pressed 
against the teeth as it is for t but comes away from them. 
(Aristid. Quint, does not mention the position of the tongue 
in pronouncing 6, and is therefore not referred to on this point.) 
Now is it a correct statement to assert that in pronouncing 
the spirant th as in English (and modern Greek) the tongue 
" is not pressed against the teeth, but only brought near " ? 

We quote the words of two English writers on the subject, 
of whom the one has made Phonetics a special study and is 
generally taken as an authority. T. Walker, Principles of English 
Pronunciation says : " th in ' think ' and ' that ' are formed by 
protruding the tongue between the iore-ieeih., pressing it against 
the tipper teeth and at the same time endeavouring to sound 
the s or z ; " and Prof. Sweet, Primer of Phonetics, p. 79, 
says of th in " thin " and " then " : — " Certainly the most 
distinct form of these consonants is that produced by placing 
the tip of the tongue firmly en the hack of the upper teeth and 
forcing the breath partly between the interstices of the teeth, 
partly between the sides of the tongue-tip and the surface 
of the teeth ; but they can be — and often are — formed 
by bringing the tongue against the gums without touching 
the teeth.'' 

Thus Walker agrees with Dion. Halic, and Sweet in his 
description of the "most distinct" pronunciation of th also 
agrees with Dion. Halic. in every point, whilst the rest of his 
definition answers to that of the Scholiast. 


It is then clearly wrong to say that the description of Dion. 
Halic. could not be applied to d pronounced as a spirant, seeing 
that it exactly resembles that which Sweet gives for the pro- 
nunciation of our English t?i, ; for we do not think that the 
meaning of 'iireira in Dion. Halic. is to be so forced as to imply 
that the first action, that of closure, must necessarily be at an 
end before the second begins. 

Briicke (p. 53) says of the hard English th and Mod. Greek 6 
" that it is of no consequence whether the tip of the tongue 
lies between the teeth, is pressed against the lower teeth, or 
lies just behind the upper teeth . . ." and then he proceeds in 
words which might almost be a free modern translation of the 
second half of Dion. Halic.'s sentence on t Bd: "Das wesentliche 
fiir diesen Laut ist, die Zunge mii den oberen Schneide?dhnen, 
und zwar mit ihnon allein die Enge hildet uiid durch die Enge 
wird der Luftstrom hervorgetrieben." 

The description given by Aristides Quint, for t S d is vague 
and does not mention the position of the tongue, but only that 
"the teeth must be a little apart and the tongue hurls forth 
the breath, so to speak, through the opening thus formed." 
This is not at all exact but in general meaning seems to be 
identical with Briicke's description. 

Add to the words of Dion. Halic. the words of Arist. Quint, 
that for the Saaii, 6, the breath must be brought from the larynx 
and we obtain a perfect and complete description of our pro- 
nunciation of spirant th. Thus the argument that 6 pronounced 
according to Dion. Halic.'s description must certainly be an 
explosive, not a spirant, entirely falls to the ground — because 
from it d might be either. 

We pass on to 0. Here Dion. Halic. and Aristid. Quint, write : 
Toy a-TOfiaTOi; •jriea&ivTO'; TO-.-irvev/jLa Xvar) top Bea-fiop avTov 


and TO, fiev "jt ^ (j) Sia t&v j^eiKiwv rjj^jUTai, (jlovov tov irvev/MaTOi; 
TTjv efj.(f>pa^iv avTOiv Kara /lecrov ex^ia^o/iivov, respectively ; 
whereas the Scholiast on Dion. Thrax says that for <f) dvoiyo- 
fiivcov Se T&v ')(6ike(ov iravv Kal irveviiaTO'i ttoXXoO i^covra 
eK^Qyveirai rb (f), — and here the objection urged against a 
possible spirantic pronunciation of is, that the lips are not 
closed in producing /. They are, it is true, not closed in 
producing our dento-labial /, but even the Modern Greek 
does not correspond exactly to our "/," and in ancient times, if 
it was a spirant, it was presumably bilabial. 

The descriptions given by Dion. Hal. and Arist. Quint, of 
the manner of articulating ^ are certainly in every point 
correct, if ^ was an aspirate ; at the same time it must be 
remembered that these definitions may have been inaccurate 
and may rather denote the general characteristics of each group 
of three letters than the little differences in the peculiar 
method of articulating each individual letter. 

If, on the other hand, ^ was a spirant, not an aspirate, can 
their definitions be accepted as in any way applicable to a 
bilabial spirant? Dion. Hal. uses the expression to irvev/jLa... 
Xvay TOV Secr/Moi' and Arist. speaks of the ef^^pa^tv rSyv 
■X^eiXeoiv, both of which indicate closure of the lips. According ^^ 
to Sweet (p. 30) no spirant is pronounced with closure or 
e/j,(l>pa^i<;, whereas others {v. Ebel. K.Z. xiii. p. 265) say bilabial 
spirant (f) is pronounced by keeping the lips closed as for tt and 
then blowing through them. 

Some of the Greeks themselves too say that they press the 
lips quite as closely together in pronouncing (j) as in pro- 
nouncing TT. 

The scholiast's description, on the other hand, is most un- 
satisfactory — the lips are never " wide open " in producing (f>. 


be it pronounced as aspirate or spirant, unless he means that 
after the breath has gone forth, the lips are forced open, but 
then that would apply equally well, or better, to the explosive it. 

Lastly, we come to ^y. In the pronunciation of this letter 
"the tongue rises to the palate near the throat," Dionysius 
Halic. says, as it also does for k and 7 ; he is thus evidently 
here describing a palatal k 7 and pj;. — Whether aviaTafievri<; 
7rpb<;. . .here means " is raised to " in the sense of " being pressed 
against " is, we venture to think, doubtful — it may do so and 
thus Blass evidently interprets it, for he argues that from this 
description complete closure of the vocal passage was required 
for sounding ^, but this is rather forcing the meaning of 
avia-Tafj,€vri<; tt/so? which seems to denote simply approximation 
of the tongue to the palate and to leave indefinite whether 
partial or complete closure took place. 

There is not much difference between Dion. Halic.'s definition 
of the palatal letters and that of Sievers who (p. 53) says : 
Unter Pa,latalen verstehen wir die durch Articulation des 
mittleren Zungenriickens gegen den harten Gaumen gebildeten 
k-ahnlichen Verschlusslaute und die diesen entsprechenden 
Spiranten e.g. ■x^. 

Certainly the avia-Ta/xeiir]<; 7r/3o?...of Dionysius Halic. is so 
vague that, as he gives no further explanation of his meaning, 
we are left quite in the dark as to whether he really intended 
to denote complete closure by these words and to include all 
three letters under explosives, or whether both 7 and x '^^ ^^^7 
Y were already spirants — if we had no further explanation of 
the nature of k and spirant ch from Sievers than the sentence 
quoted above, and no other grammarians to refer to, it would 
be difficult to tell wherein as regards the position of the tongue 
the difference between Germ. Jc and ch lay. 


Aristides Quintilian says of this group : to. Se ^;\;6tTat, rrj<; 
fiev irapeia<i VTroaaipovar)<;, rov Be irveviiaTO's payoaito^ Kac €i? 
TrXaro? 7rpo'i6/j,ivov. This description is almost useless as he 
omits all mention of the tongue, -which is the most important 
factor in the sounding of these letters, and what he does say is 
hard to interpret. That the cheeks should "grin a little" 
recalls the advice of Modern Greeks, who when trying to explain 
to foreigners how to pronounce their spirant pj; constantly tell 
them to pull down the corners of the mouth. 

The Scholiast on Dion. Thrax gives us a more accurate de- 
scription of the letter ■)(^ than he does of any other, as, after saying 
that for K the tongue is pressed against the palate, he goes on 
to say of ■^, Trj<; yXd}TTr]<; firj Trpoa-TriXovfiivr]^ firjB' oXax; 
avvaiTTOfiivrj^ t«5 ovpaviffKfo, which is of course the correct 
description of a spirant ■^. 

We have thus endeavoured to show, firstly, that the remarks 
of the grammarians must be used with caution, as they are very 
incomplete and we do not know definitely in what sense they 
use terms they employ, nor do we know for certain when 
Aristides Quintilian for instance lived, which is of importance 
for the value of his evidence ; and, secondly, that if the remarks 
are carefully examined and weighed, compared with each other » 
and with the remarks of modern grammarians, and translated in 
an impartial way and suitably to the context, they are at the 
best very hazy, indefinite and often incomprehensible. 

In fact it is useless to try and derive conclusive arguments as 
to the specific nature of the letters involved from these sparse 
remarks of the grammarians, whose classifications may have 
been no less, or even more, faulty than those of some English 
and German grammarians of the present day. 




Apart from the grammarians we can glean very little from 
Greek writers to help us in deciding the probable pronunciation 
of'x^Kpff, but what there is we will now briefly examine : — 

In Aristoph. Thesm. 1. 1183 S. a Scythian is introduced who 
pronounces the ■)(^^0 oi the Greek words as ktt t, for instance 
TvyaTptov for Ovydrpiov, airoTpeKe, iXairpos, Kaplev, KeTraXy^ 
and so on. 

From this barbarian's manner of imitating the aspirates Curtius 
and Peile draw the conclusion that % </> ^ were certainly still 
real aspirates in Aristophanes' time, but we do not consider this 
conclusion justifiable. 

Let us compare the pronunciation of this foreigner of ancient 
times with that of foreigners of present times and see whither 
this will lead us. 

The French pronounce the Modern Greek spirants '^^ and hsk 
and T, the Germans likewise say t or d for and, what is more 
remarkable, they nearly all say k for ^^ in spite of their having a 
spirant 'ch' in their own language, and the English also 
generally say k for ■x^; the is of course pronounced by all 
these as f. 

Similarly French and Germans say t or d for English spirant 
' th,' the Russians say t or f for 0, the Arabians of Egypt and 
Syria pronounce the hard Arabic ' th ' as t, and the Slavs and 
Lithuanians represent Modern Greek ^ and f by p, and German 
ch by k (c/, Blass, p. 102). 


Thus then we see that in very many cases the spirants of 
a language are pronounced as the homogeneous tenues by 
people of other nations, and that the present spirantic ;y; 9 ^ are 
usually pronounced as kttt by those who cannot pronounce 
them correctly. This being so, is it a just deduction to make 
that because the Scythian in Aristophanes said tt k t instead 
of X ^, these latter must have then been true aspirates, when 
we see that the French, German, and Slavs of to-day say tt /& t 
instead of spirant ^ x ^ ? It by no means seems reasonable to 
make such a deduction in the face of the facts we have men- 
tioned, as it might with equal justice be argued that since 
spirantic (/> x ^ are often in the present day spoken as tt « t by 
foreigners, then the Scythian's saying tt /c t for the ancient <^ x ^ 
must prove they were then what they are now, that is, spirants. 

However it is unnecessary and impossible to arrive at a 
definite conclusion as to the value of ancient ^ x ^ from these 
Scythian mispronunciations — for, if they were true aspirates 
and he had none in his own language, he would naturally have 
spoken them as tenues ; or again, if they were spirants and he 
had none similar to them in his mother-tongue, he would 
probably have spoken these too as tenues, as is so constantly 
being done at the present time. And, just as Aristophanes 
ridiculed the Scythian in his play, so the foreign pronunciation 
of the Modern Greek spirants is often now introduced on the 
stage and certainly has a very comic effect. 

An argument often advanced by the supporters of the 
spirantic pronunciation of ^ x ^ is> ^"^^^ if these letters had been 
real aspirates, then the difference in sound between them and 
the Scythian's pure tenues would not have been noticeable or 
peculiar enough to have aroused the mirth of a mixed Athenian 
audience. It does seem doubtful whether it would have been 


distinct enough and yet the probability is that it would have 
been, for we find that the Germans laugh at the Czechs of 
Bohemia when the latter pronounce pure tenues in German 
words — and this habit is often made the subject of jokes in 
comic papers — and vice versa Hungarians laugh at Germans 
when they aspirate the tenues of Magyar words. Thus this 
Reuchlinian argument falls to the ground, but so also does that 
of the Erasmians who judge from the Scythian's mistakes that 
4>X^ must have been aspirates, and thus we are reduced to 
deciding that this passage of Aristophanes is of no assistance to 
us at all. We must also remember to leave a little margin for 
exaggerations in the caricatures of a comic poet, and can there- 
fore not rely sufficiently upon him to assert that all Scythians 
always said tt for <f) and 'ApTefj,ov^t'a for ' Aprefitcria. 

The only other passage we need mention is Plato, Cratylus, 
427 A. hia Tov <f> Kol Tov yfr Koi tov a- Koi tov ^, on irvevfia- 
TcoSrj T^ ypafi/iara, iravTa to, Toiavra /Me/JfifirjTat avroi<; 
ovo/xd^eov, olov to ifrv^pov koi to ^eov /cal to aeieadat, Kal 6 
creia-fibi;, Kal OTav ttov to ipvcratSei; fitfifJTai, iravTa'^ov 
evTuvda o)? to iroXii to, TOiavTa rypd/j,fx,aTa iirKpipeiv <f)aiveTai 
6 TO, ovofiaTa Oe/jLevo<;. This Jowett translates as follows : 
" And there is another class of letters, ^ yjr <r ^, oi which the 
pronunciation is accompanied by great expenditure of breath ; 
these are used in the imitation of such notions as '\jrvy(p6p &c., 
and are always introduced by the giver of names when he 
wants to imitate what is windy (^uo-wSe?)." 

This Raumer takes as a positive proof of ^ having in Plato's 
time been a spirant, whUe Blass thinks it should not be so taken, 
and seems to find an explanation of the epithet TrvevfiaTwBrj 
in the fact that " a- has an aspirating power," and translates the 
sentence " ^ i|r cr f are letters with a strong breathing." 



There can, however, be no doubt that Plato did not intend to 
convey any such notion by his word Trvevf^aTtoSr), if the passage 
is read in connection with the context, and especially with the 
next sentence, which says "these letters are used to imitate 
what is ' windy ; ' " hence TrvevfiaTcoSTj must mean, as Jowett says, 
that they require a great deal of breath. Placed in juxtaposi- 
tion with simple and compound sibilants as it is here, and then 
further being designated as one of the letters which is used to 
imitate what is "windy," certainly seems as if it would 
answer better to this description if =/than if it were =p +h, 
and it would also more fitly be ranked with sibilants. Without 
further evidence to confirm it these few words of Plato's are 
not sufficient to settle the fact of cj) being a spirant, but such as 
they are, they certainly seem to indicate that it was such and 
no longer a real aspirate. 



There are various isolated facts connected with the ancient 
orthography which are taken as confirmation of the aspiratic ^ 
pronunciation oi ■^cjid, and amongst them is this one : that 
"those Greek races which did not possess the non-Phoenician 
symbols cj) and x, in tliose cases where they were not satisfied 
with the simple tenues, adopted the writing IIH, KH." Among 
the races who were content with the simple tenues were the 
Cretans, who wrote tt and k for and 'x^ — e.g. icpifmra = 
'X^prifiaTa. The inhabitants of Thera and Melos, however, 
wrote IIH, KH for ^ and x '> e.g. eKirhavToi, = eK^avTou, for in 
the oldest inscriptions from these two islands and which 


Kirchhoff attributes to the second half of the seventh century 
B.C. there are no special signs for ^ i/r </> p^;, and in place of 
the two latter irh, kK are consistently written, which fact is held 
to prove that and ;)^ were on these islands at this period 
pronounced as explosives followed by an " h." It is perhaps, 
however, unwise to draw this absolute conclusion, for although 
IIH, KH are written for ;)^, TH is on no inscription written 
for 0, but always 6 or @H, and these writings for ^ and ;y; 
may, for all we can tell, be merely a clumsy attempt at 
representing the spirants for which, owing to the then in- 
complete state of the Greek alphabet, they had no special 
signs, whereas the aspirate or spirant 6, for which they had 
a Phoenician sign, is never resolved into its component parts. 

These old writings can no more be used as incontrovertible 
proof of the aspiratic nature of <^ and ;j^ than the English 
writing " th " and the German " ch " could be used hereafter 
to prove that these " th " and " ch " were spoken as explosives 
followed by an " h." 

But it seems much more strange that, if ^ ;)(; were really 
spoken as explosives followed by an "h," the Greeks should 
have felt the need of new signs to express them — if IIH, 
KH exactly denoted their phonetic value, why was this writing 
abandoned and another adopted which did not clearly represent 
their sound ? It seems a needless and rather senseless inno- 
vation on the assumption that and ;)^ were aspirates; if 
they were spirants, it is easily comprehensible. 

Still, even ii ^ 'x^ 6 were real aspirates, the necessity for 
the introduction of these signs may very well have been that 
the value of H itself had changed, so that it now denoted rj 
and consequently HH, KH could no longer be used to denote 
p + h and k -f- h. 

E 2 


Again, that 6 was au aspirate, and further that a need 
was sometimes felt to distinctly express the breathing which 
followed the tenuis, is said to be shown by the writings 
0hapv/jLaKha, dhapvfia^ho^ on the old Thera inscriptions, 
I.G.A. 444,449. Similar perhaps is ^hpahcrov from Naxos, 
I.Q.A. 407, if the <J> is correct, but it is doubtful, and even so 
the h here possibly belongs to the p, not to the ^. We must 
notice also that it is only twice that the " spiritus asper " is 
written after Q on the Thera inscriptions, on the others it is plain 
6, e.g. OeoOefMo';, ih. 469 ; 'Op6oKKri<;, ib. 451 ; ^eros, ib. 456, &c. 

This insertion of the " h " after 6 may denote its aspiratic 
nature, but it may also have no real meaning. The spiritus 
asper was used somewhat loosely in Greek ; in Old Attic it 
is sometimes inserted between vowels, it also appears after 
initial \, p, and p, e.g. Xheaiv, fiheydpei,, its insertion in these 
places is explained by Blass (p. 88) as due to the fact that 
\ and p, when initial had their fullest, and not a weak, sound, 
as /i especially had when medial. 

This same explanation might stand for the aspiration of 
initial 0, or otherwise the insertion may in these two cases 
very likely be an error. 

Brugmann (GreeJc Grammar ^, p. 65) says that the insertion » 
of the "spiritus asper" after the liquids showed that they 
were voiceless in these cases. 

Another circumstance generally taken to prove that ^ and </> 
were real aspirates up to the time of Eukleides at any rate, 
is this : that before the introduction of the symbols f -yjr, the 
inscriptions from Attica, Styra in Euboea, and Cumae in Italy 
employed -x^a; ^o- for them, e.g. ehoxosv, ypdcfxTat, C.I.J. 21 ; 
Cumae K\e<j)a€, I.G.A. 524 ; Styra Mo^o-%9, but also xapoTri ; 
whilst in Amorgos and Thera they employed «cr, tto-, which 


latter is undoubtedly the most correct manner of resolving the 
sounds ^, 1^, e.g.. Aafiira-ayopeio, Bechtel, Inschr. d. ion. Dial. No. 
29, from Amorgos ; 'VeKo-dvwp I.G.A. 451, Ua-riv ib. 461, from 
Thera. Now what do the writings p(;o-, <^cr mean ? That ^=k + 
A. + ? and -^=77 + ^+9 or that ^ = cA + 9 and 1^ = / + 9 ? 
If it means the former the h must certainly in this position have 
been all but inaudible, and then with the extremely slight 
difference in sound there can only have been between such a 
kTi, Trh and pure «, tt it is curious that plain ira-, Ka are never 
inadvertently written for ^o", ')(<j- in the pre-Eukleidean Attic 
inscriptions, especially in such a word as e'f , which even stone- 
masons probably knew was another form of sk and yet we never 
find 6'«9 but e;j(;9. To say that kJi, •jrh were liable to be heard 
instead of k, it ^because ^ was a jpdfifia irvevfiaT&Se^ is not 
sufficient explanation — we have no authority for saying that 
a lypdjifia irvevfiaTMSef means a " letter which has an h as 
one of its component parts." 

Another fact which militates against our rashly assuming 
that the % ^ in ■x,a; ^o" were aspirated tenues, and not spirants, 
is that on the archaic inscription of the sixth century B.C. from 
Naxos {I.G.A. 407) neither ko- nor ;i^o- are employed for ^ but ha- 
(gS) : thus e\-(To-xp';, 'Na\- clov, ^ypdyaov; while on a later 
Naxos inscription of the fifth or beginning of the fourth century 
B.C. {I.G.A. 410) %cr (X^) is used, e.g. ' AXxcTrjvtop, Na;\;o-to9. 
These two facts when looked at side by side at once suggest, 
firstly, that the first component in the letter f was a spirant at 
least to Naxian ears, if ha could be employed for |; and, 
secondly, that ;)^ must have been a spirant also, if it was after- 
wards used for h in ')(a- as = f . 

Unfortunately i/r or a substitute for it does not occur on the 
ancient Naxian inscription, ■ 


Another ancient inscription of the fifth or fourth century B.C. 
also found on the island Naxos which has Aa)po(f>ea for AapoOea 
(the <f> is very distinct), cf. Bull. Oorr. hell. ix. p. 495, makes us 
think that and must have been spirants in Naxos by the 
fourth century B.C., and if these two letters were spirants then, 
that would make it still more probable that x should have 
been one too. 

It is quite likely that some races heard a hard explosive as 
the first component in | and yjr, and others a softer, spirantic 
element and therefore .represented these compound sounds 
differently. In any case the name •n-vevfiarmSe'; would apply 
equally well to ^ and ■s]r ; as Plato includes o- under that class 
likewise, the mere fact that a was one of the components of ^ 
and yfr would suffice to explain why he puts these two letters 
under the same category. 

With this ancient Greek method of expressing ^ by ^a; where 
the X i^^y have been a spirant, we can compare the German 
manner of expressing the £c-sound by chs, where the " ch " is not an 
aspirate but a spirant : e.g. wachsen = waxen, Fuchs = Fux. 
In the German ch the explosive is not distinctly heard, it is a 
bond fide spirant and yet this spirant ch -\- s is consistently 
written to express the sound x or ^. This parallel removes the 
difficulty some feel to the possibility of •^ having been a spirant 
when used with a to express the sound |. 

Chap. VIII. 



y Evolution of the Phonetic Laws. /Ancient Greek ttt, kt, <^Q, xO, becomes 

<^T, ;^T in Modern Greek 




a few instances illustrate the 
tendency in Ancient Greek 

/In Modern Greek a surd spirant often /before cr — change general in the 

becomes tenuis before and after a- 

vernacular, with few excep- 

in both Ancient and Modem the change in 
the case of the double tenues predominates. 

in Ancient Greek find the tendency in Doric 

after a — double forms in both Ancient and Modern difficult to explain according to the 
/ aspiratic theory. 

An Modern Greek the nasal before an 
Aspirate is often omitted 

Ancient Greek examples of this 
omission, which is part only 
of a larger modification 

both Ancient and Modern omit nasals before 
tenues, medials, and betw-een vowels. 

In Modern Greek 6, .<^, y(^, sometimes ^rather the .exception than othei' 

become tenues after p 
Chap. IX. 
The History of Interchange. ^Aspirates with Aspirates y{a) cf, for t), (J) for .^ 

/(«) X for e, (*) e for X 

/change of A. into p before <^ disproves existence of any law to this effect. 

/{a) X for <^, (b) <j> for x 
/Aspirates with Mediae y(a) S for 6, {b) d for S 

/(a) fi for 4,, (J) <l> for ^8 

/Aspirates with Tenues 

/Two consecutive Aspii'ates 
for Tenues 

/(a) y for x, (b) X for y 

y{a) T for 6, (b) for t 

/(a) TT for <j>, (b) (f> for ir 

/{a) K for X, {f>) X for k 

X6, i>0 for KT, ITT 

{a) rare in Ancient (except Aeolic) and Modern ; (6) occasionally in Ancient 

and Modern, 
(a) rare in Ancient and Modern ; (&) neither in Ancient nor Modern. 

(ct) and (6) occasionally in Ancient and Modern. 

(«t) rare in Ancient and Modern; (J) occasionally in Ancient, never in 

{a) i^xre in Ancient and Modern ; (i) one instance in both Ancient and 

(a) and (6) rare in Ancient and Modern. 
(a) and (b) frequent in Ancient and Modern. 
(«.) and ('&) rare in Ancient and Modern. 
(«) frequent in Ancient, rare in Modern; (h) frequent in Ancient and 

a few cases in both Ancient and Modern. 

/Aspirates with Sibilants /(a) cr for 6, (b) a-a for 0, (c) for o- 

o- for X I iiot found in Ancient, frequent in 

y{a) Ancient Laconian, Modern Laconian. 
/(6) Ancient Ionian, Modern not found. 
/{c) Ancient rare, Modern not found. 


Modern dialects 


Chap. X. 
Elision and the Spiritus Asper. 

/In the various dia- /Aspiration yjyro Aspir. regular a.spiration in Attic, etc., proves ; A.spir. =Ton. + Spir. Asp 
lects ■ • f ■ •'V- 


/pro S'pir. /no errors, such as ttA written for ^, kc. 

/preceding tenuis affected in spite of the weakening of the Spir. Asp. 

/Can aspirates exist in a language when there is no initial h 1 | other languages referred to as examples. 




But perhaps the most satisfactory and conckisive argument 
as to the nature of the Greek aspirates is based on a thorough 
investigation of the Greek written language itself. In order to 
conclude whether any considerable modification in the pronun- 
ciation of certain consonants has taken place during the progress 
of a language we must observe, not only to what extent the 
phonetic laws have been modified, but also what is the nature 
of such modifications. For certain modifications are natural to 
the growth of every language. Thus the tendency towards 
greater ease of pronunciation is natural to every language, and 
may be said to be, so far as it is possible, a constant factor ; all 
modifications, therefore, based on the phonetic laws of assimila- 
tion and adaptation we must expect to find increasing and 
expanding in the course of the development of a language. So 
also we must look for the substitution of an easier sound for 
one more difficult of articulation, and thus it is that we expect 
the interchange of consonants of the same class, that is of those 
produced by the same vocal organs, according as greater or less 
exertion is required in their articulation ; and not only this, but 
we expect the consonants of one class to be substituted for those 
of another in the different dialects of the same language, accord- 
ing as there is a local preference, indicating greater facility, for 
sounds produced by one set of vocal organs rather than by 
another. Thus the lonians retained k for tt, e.g. Kolo<i for 
TTOto?, the Dorians t for a, as tu for av, irkaTiov for -TrXrja-iov. 
Another modification of sound proceeding from the tendency 


towards ease of articulation is the entire falling away or draf- 
ting of sounds, thus we have a-<l>aKk(o and /alio, .and vice versd 
olvo<; and vinum, and for final sounds the omission of m, on early 
Latin inscriptions, and in the later stages of the language, 
whilst in contemporary Greek the vulgar pronunciation omits 
e.g. the v of the ace. sing., saying avOpwiro for avOpairov, 
also that of the neuter termination, thus to hevhpo. Medial 
sounds are similarly dropped to avoid difficult combinations of 
sound, or simply to reduce one in itself not difficult to what 
requires still less exertion. 

Given two stages of a language, therefore, separated from 
each other by a very considerable interval of time, we must not 
forget to make allowance for the natural expansion of phonetic 
changes based on the tendency above illustrated, and we must 
expect proportionate modifications of the original phonetic laws. 
We must, in other words, be careful not to interpret as a change 
■ of phonetic law what is only a modification, or a further applica- 
tion or extension in a parallel direction, due to the course of 
natural development. 

We should also guard against assuming that we can lay down 
an absolute standard as regards the comparative ease of pro- 
nunciation of certain sounds or combination of sounds. For 
one sound may not only in a particular case be relatively easier 
to pronounce than another, which theoretically analysed requires 
no greater effort to articulate, but any certain kind or class of 
sound, which theoretically analysed is ascertained to require 
less effort for articulation than another, must not consequently 
be regarded as relatively easier of pronunciation. For, if we 
conclude in this manner, we are leaving out of sight the con- 
sideration of physical, and oftentimes theoretically unaccount- 
able, national and local preferences. 


Thus we are not justified in concluding that because we 
cannot pronounce such a combination of sounds as " prst " and 
" krk," therefore a Slav cannot do so, or, to come nearer home, 
that, because certain combinations of consonants are so difficult 
for us to pronounce as to be quite impossible, therefore a Welsh- 
man cannot easily pronounce them. 

Finally, much precaution is necessary in avoiding an assump- 
tion that there is anything of the nature of absolute as regards 
euphony. It is better, indeed, not to argue at all as to what 
sound may be said to be more euphonious than another, than to 
risk losing sight of the fact that what may sound euphonious to 
us, may not sound, or have sounded, so to others ; and vice 

The caution, however, that is necessary in arguing from a 
certain standpoint as to ease of pronunciation and euphony does 
not render these tests altogether inapplicable, in so far as from 
a careful examination of the phonetic laws of the language under 
observation we are enabled to gain a more or less accurate idea — 
this varying with the nature and abundance of material at our 
disposal — of the relative standard of either as illustrated in that 
particular language. 

Bearing these preliminary and necessary observations con- 
stantly in mind, therefore, we will now proceed to review some 
salient points in Modern Greek phonetics, and examine these 
carefully by comparison with the ancient language. 

If in doing so we think we can show that, apart from what 
may be considered natural development and extension, the 
original phonetic laws have been preserved, and the same inter- 
changes of sound take place now as took place in the earlier 
stages of the language, we shall conclude also that those sounds 
to which such phonetic laws and interchanges apply have 


suffered no radical change in the interval, so that their present 
would, for all practical purposes, represent their ancient pro- 

In the first place, then, one of the most striking features of 
the modern language is the dislike evinced to the juxtaposition 
of two surd spirants or two tenues ; and to avoid such a combina- 
tion in the case of two spirants coming together, the second one, 
0, is generally changed to the tenuis ; and in the case of ttt or kt 
the TT or K is changed to the corresponding spirant ; thus for 
(fiddvco we have (pTava, for ixOe<;, epjjres, and for KXeTrrri';, 
/cXicfiTrji;, for vvKra, vv'^ra. This rule is, however, not in- 
variably observed as ^0 is still frequently heard : avaTrav0rjTe = 
dvaTra(f>0rJTe, for instance, is never, to our knowledge, pro- 
nounced avaira^rriTe, nor does it hold good as yet for the 
combination -kit- which does not generally change, thus 
eKTropevo/iai, &c., do not become i'^ in the ordinary 
language, though in the dialect of Trebizond we already find 
instances of this change, thus d'^Travm for eKTrriyaivco, 
d')(pTapd^a) for eKairapda-ao). We can, however, diregard these 
few exceptions as they do not invalidate the general rule. 

The frequency of the forms with <^t, 'xj is adduced as being 
an example of a new phonetic law, unknown to ancient Greek, 
and which therefore justifies the conclusion that the sounds % ^ ^ 
must in the ancient language have been fundamentally different 
from what they are now. 

Such forms are frequent it is true, but have we not here an 
instance of development and extension rather than of radical 
difference in the phonetic law ? Of development very consider- 
able indeed, to judge from the few precedents extant in the 
ancient language, but still of development ; and we must not 
forget that it is not only from the qwintity of examples that we 


must draw our conclusions, especially when, as is the case 
in Greek, we are the less justified in so doing because of the 
small total remnant of the language of the uneducated and of 
every-day intercourse as compared to that of the educated and 
literary world. 

Bearing, then, this caution in mind, we turn to the ancient 
language to see if we can find anything to justify us in thinking 
that this modern feature is not a change, but a development of 
a law already anciently existing. 

There are apparently no traces of a writing (pr for ^6, or of 
0T for TTT, but of X''' fo'^ "'^ there are three, or maybe four, 
examples extant and perhaps one of y(^T for '^0, and these 
examples, few as they are, are most valuable as precedents of 
the modern usage. They are the following : 

from Attica 'Evray^ro'; = evTaKTo<;, Roscher, Gurt. Shod. i. 

2, p. 81. 
from Chios e'p^; t&v = i/e rmv G.I.G. 2241. 
from Sagalassos Karej^ravev = KareKTavev ib. 4377, 5. 
and from Elis eV Ta^rd = iv TaKrrj (?) I.G.A. Add. 113, 
c. 2. 
The second and third examples belong to the Roman period, 
while the last one is probably not much later than 572 B.C. (v. 
Roberts, Greek Upigraphy, p. 298). 

Thus we have three definite cases dating from pre-Christian 
times of j(t being written for kt. The last example, which 
Bucheler takes as equivalent to eV TaxTrj (cf. Bhein. Mus. 
xxxvi.) and which would, if his view be correct, make the 
fourth instance of 'x^t for kt, Comparetti (cf. Journal of Hellenie 
Studies ii. p. 373) interprets differently, for he takes the 
original evrdyfrai as standing for ivrdy^Oai, i.e. eVreraj^^at with 
an omission of the reduplication, and r an error for 6. If this 


be so, this word would be a solitary example from the ancient 
language of 'yr for ')(jS. Blass corrects it into ev ravrai (v. 
Roberts, p. 369). Of these instances the contemporary language 
shows not only development by the frequency of such forms, 
but extension likewise by the application of the same change 
to two successive spirants, and also to the group ttt. Thus we 
have firstly -xj ^^^ ^''' instead of the two aspirates ■)(j9, ^6 in such 
cases as : i'irKi')(r'r}Ka = iTr\,e')(6r]v, e-^Te'; = ;\;^69, o'XTpo^ = i'^Opo';, 
(j)Tdvo) = <f)0dvco, (j)rrjv6<; = evdrjvoi, iypd(f)Tr]Ka = eypdcfydrjv, etc. 
and secondly 'xj, <^t, instead of the two tenues kt, ttt, examples 
of which abound, such as : dSpd^rt = drpaKToi;, dxriSa = dKTl<i, 
o^T«, vvyra, crrdxTr), yaXa-yTl^co, /S/ae^^ro?, ai/ojT^ros, ^pd^rf] = 
tppdKTrji;, <f)ovj(Ta = irvKTi}, airpaxro^, i^ TrjOrji; (Trebizond), 
(also initial) xjfjfia, ^^rtfo), xjivi = KTel<s, x'^VTrijcre (vernac.) ; 
and again, (pTvco, <^Taixp<s, <f)Tepva, ^racjo) = TTTaiw, (jiTepo = 
inepov, <f>Tapvt^o/j,ai = Trrepvi^ofiai, pd<f)Tr)<;, /cXe^Ti??, 6</)Ta, 
/3a(j)Ti^(o, Tre^Tto, vi(j)Ta), K6<f>Tco, d(npd(^T(o, d(j}TCO, ddacjiTO^, 
&c., &c. An explanation has been suggested, according to 
which we have an extension by analogy in the contemporary 
language of the change of into t after a- already so prevalent 
in the ancient language. So that would subsequently have 
been changed into t not only after a-, but after ^ and (f> also, , 
and then the familiarity of the sounds 0t, y(T would have led 
to the frequent change also of tt and k into (f> and ^ before t. 
The objection to this explanation is that it assumes the priority 
of ^0, '^0 becoming ^t, yj over that of ttt, kt being changed 
to 0t, yr, and the considerably more numerous instances of the 
latter change in the contemporary language at least seems to 
argue against such a priority. The latter change is actually far 
more common in contemporary Greek. As it happens, three of 
the four examples surviving from the ancient language are 


instances also of the change of ttt, kt, not of (^6, yd, into 0t, 
P^t; but this, since it may be merely accidental, cannot be 
taken as evidence. Considering, however, the difficulty which 
prevents our accepting without hesitation the explanation 
adduced, we prefer to consider the present combination of 
spirant and tenuis a development and extension of what had 
actually begun in the ancient language, but of which we have 
very few instances remaining. 

Again, it is laid down as a distinctive law of Modern Greek 
that " (J does not admit of a surd spirant either immediately 
preceding or following it " — here, likewise, we must investigate 
whether this is quite a new law, unknown to the ancient 
language, or whether it is not rather the natural development 
of a tendency, the beginning of which can be seen not only 
in other ancient dialects, but even in Attic. 

We will first take the case of a spirant immediately ■preceding 
<T ; this in the popular language is now generally changed into a 
tenuis. The most numerous examples of this change are found 
in the future and aorist endings of verbs in -avos and -evw, which 
instead of -avaco, -evtrio, &c., are generally spoken as -dyjrm, 
-iyjra), &c. (-avaeo, -evato being = -d^am, -e<j}(7Co), thus KKavaw 
becomes Kkd^jrco and i^aaiXevaa, i^aalXeyjra ; further for 
Kddiae, Karae is frequently heard. 

On the other hand, a word in constant use is dtjxre, a 
contraction for d<f)r)a-e or a^crov, and this is never changed 
into ay^e, although the spirant directly precedes the a. Simi- 
larly iravata, e-rravaa, &c., are quite as often, if not oftener, 
spoken as Tra^o-o, eira^a-a, &c., than as Trd^frco, eira^yp'a, &c., and 
the same can be said of the forms of Trto-reuo). 

Not only has no definite law been established in Modern 
Greek by which o- does not admit of a preceding surd spirant. 


but we find the tendency now so fully developed originated in 
ancient Greek. Thus in Doric we have ■^e (Theocr. iv. 3), 
and yjriv for o-^e and a^tv respectively, these latter haying by 
metathesis become 0o-e, (fxrtv, and then yjre, ■\jnv. 

Further illustration of the fact that this tendency already 
existed in ancient Greek is afforded by the forms ■y^iat,';, 
•ylrivofjiai, ■\jri,vdSe^, Siyfrdpa, /j,6po^o<;, e^iarov, and othefs given 
by Hesychius. These are respectively equal to ^dierif, <pdivofj,ai, 
&c., and can only be explained from a spirantic pronunciation 
of (f> and 0, so that to arrive at the forms in question the 
following successive stages may have been passed through, 
e.g. <l)diai<; ( = fthisis) = ^a-i(Ti<; = ■y^LuK; ; Meyer {Greek 
Grammar^, §§ 209 and 250) gives the stages as probably ird — 
Trth — yfr. We thus have exact counterparts to the modern 
writing ■yjr for <j)a: 

We will next consider the dual forms of words written with 
either aspirate or tenuis after the letter o-. Such forms are 
numerous, both in Ancient and Modern Greek, e.g. a-y^^eXi'; and 
a-KeXi<; in Ancient, o-^v/Ji? and <TnTvpl<s in Ancient and Modern, 
da6evri<i and daTevij'i in Modern. The question for us to 
consider is whether the almost consistent pronunciation of o-«, 
a-T for ax, oS (but not a-0 for o-tt) in Modern Greek is indicative 
of a new phonetic law unknown to classical Greek, or whether 
it is a continuation and development of a process which had 
already then begun. Blass (p. 103) lays down as one of the 
distinctive phonetic laws of Modern Greek that " a does not 
allow of a surd spirant immediately following it." And 
Curtius and Roscher teach that of the duplicate forms those 
written with an aspirate in Ancient Greek are due to the 
aspirating influence of the preceding o-, and that they are 
later forms than those written with a tenuis. This theory 


seems to have been generally accepted, and according to it 
the adoption of ctt, o-k for aff, a-^ in Modern Greek would 
certainly be a distinct reversal of the ancient law. 

But are the forms written with tenues older than those with 
aspirates? And has the aspirating influence of o- been fully 
established ? Bezzenberger (cf. Bzzb. Beitr. vii. 63 if.) pro- 
pounds another theory in which he is supported by Gustav 
Meyer (v. Greek Grammar, 2te Aufl. p. 207) and Grassmann, 
and that is that the forms with aspirates are the older and 
then in many cases the aspirate has been changed into a tenuis. 
The aspirating influence of <r has not, he says, been proved, 
as in many cases adduced to prove it the derivation of the 
word adduced is very uncertain, in others the aspirate-form is 
recognised as quite as old or older than the tenuis-form, and 
again, aspiration is in many cases produced by analogy, as in 
Siy(^o/Mat from JSeK ; Tevj(<o from Jtvk. 

For the majority of these double forms the aspirate is due 
to an original Indo-German surd aspirate which has been pre- 
served in Greek, although very often it is represented by a 
pure tenuis in the kindred languages. Curtius, on the other 
hand, says that in these cases the surd aspirate of the Indo- 
German became a tenuis in Greek first and then became an 
aspirate again owing to the preceding sibilant. But why this 
step backwards and then another forwards to restore the word 
to its first state should be necessary, or even probable, it is 
difficult to understand. 

It seems much simpler to suppose that the Greek surd 
aspirate in these cases corresponds to the Sanskrit surd 
aspirate and was the earlier form of the word, and not only 
simpler but more in accordance with facts also, as in some words, 
e.g. \ia-<j>oi and 0-^07709 with by-forms Xt'o-Tro?, a-ir6rfyo<;, the ^ is 


recognised as being prior to, or at least quite as early as, the ir. 
a^6yyo<; is from stem spheng, whereas airo'yyo'i is a later 
form, and Xia-(f>o<; appears to be for Xtr-fo? whete perhaps the 
tj) may be taken as representing the F, and the by-form with tt 
has not been proved to be earlier. 

So too of a-K60p6^ and a-xedpoi; (a-)(eBp6<;), cr'xedpoi is un- 
doubtedly the prior form as its derivation from cr'x^ed-eiu shows 
(and the k is due to dissimilation). And yet these three words 
are constantly placed among those in which the aspirate is said 
to be due to the a. 

Of other words which were written indifferently with aspirate 
or tenuis (although generally preferring the aspirate) and in 
which the aspirate appears to correspond to an Indo-Germ. surd 
aspirate we may mention : — 


a-a-((j))dpajo<i which with a-cfidpa'YO'; (only found in com- 
pounds), a-Trapydaj and (j>dpvry^ appears to be connected with 
Sanskrit sphurg, a<l>vpL<i, a-cftvpd';, a-(j)ovBv\7], cr(j)6vSvXo<; {cf. 
(nraipco, cnreopm &c.), d-cr<l)6SeXo<; : Sanskrit sphar = whirl. 

a-'xevSvXT], (T'x^bvBaXfjLO'; or a-KivSakfi6<i, a-'^^d^a (cf. also aKsMv- 
vvfib &c.) : Sanskrit skhad = split. 

a-)(eXi<; (cf. crKeXo?) : Sanskrit sJchal. 

Besides these of Xta^o? and Xio-tto? the tt can "only claim 
relative priority." So also Boo-^o/ao? and Boairopoi;, probably 
connected with <f>ep(o. Ato-;y;\a/3tfi3 is found for 'Ao-kXtjitioi! at a 
very early date (I.G.A. 549), and as the etymology of the word 
is quite unknown, it is impossible to say whether % or « was the 
earlier ; so too ' Ka')(\aiTia)v (third century B.C.) new C.I.G. 

The probability of this supposition is enhanced by the evidence 
we have that a de-aspirating tendency after o- was at work very 


early in some dialects and even somewhat affected the Attic 
dialect ; that the inverse tendency should have existed at the 
same time in the same dialects is not at all likely. 

This de-aspiration is most evident in the Elean and Lokrian 
dialects in which ctt is constantly written for a-6, e.g. Locr. 
iXea-rai, d(pe<nm c. 431 B.C. I.G.A. 322 ; El. Tifitoa-roav, KeKoi- 
(TTav, Xva-daTM I.Cr.A. 117, 119, 121. Further Phoc. airoiroKi- 
Tevaaa-Tai and •Kapw^evkaa-rat, second century B.C. Bull. Corr. 
hell. V. p. 898 ; and from Elis again irdaKoi = 'jrda-')(ot, I.G.A. 
112, of sixth century B.C. 

from Attica Kadapi^ea-Tio, O.I.A. iii. 743, of second or 
third century B.C. 
„ ,, 'ETna-rivov, C.I.A. ii. 2683, of second century 


„ „ AaKia-revov, C.I.A. ii. 1499. 

dcTKaa-Tovi ( = da-y^darov!), G.I.G. 4255, fourth 
century B.C. 

„ „ ' AXKKTTevov, 'A6r)v. v. 417. 

„ Megara Alyoa-revlrai,, Mitth. Arch. Inst. viii. 183. 

„ Messenia dyeia-Ta, Cauer^, 47, 28. 

„ Delphi yivecTTco, dia-rmv, &c. 
Since therefore we have these proofs that not only was 
de-aspiration after <r at work more or less in all dialects but also 
that of the words written with aspirate or tenuis, the forms with 
aspirate can in some cases undoubtedly claim priority whilst in 
the others they cannot be proved to be posterior to those with 
tenues, the rational conclusion appears to be that the forms of 
words written with aspirate after a are the original and 
that the de-aspirating influence of the sibilant a produced the 
forms with tenues. Further, that this law of de-aspiration after 
(T, which bad commenced in the classical era, especially outside 


Attica, has continued to gain force but is not even yet fully 
developed, because though it holds good for o-;^;, which is now 
always pronounced o-«, and mostly for ad also, yet acf) still holds 
its own and is only in Trebizond pronounced as o-tt, which shows 
that the change already completed for a^ and ad has also 
begun for a-^. 

Contrariwise, if the combination cr^ with ^ = a spirant can 
be pronounced as we know it still is at this day, a-'x, and a-d 
with x and = spirants most probably also existed at some time 
(cr6 = sth has not died out yet entirely) or other in the ordinary 
language, v. Psichari (Mdm. d. I. Soc. de Zing. vi. 304). 
When were the % and in a-^, or a-0 pronounced as spirants ? 
We see that now in 1890 A.D. they are usually pronounced as 
tenues in these combinations and we have evidence from the 
inscriptions that by B.C. 400 they had already begun to be so 
pronounced, so that if at that date ;)^ and d were real aspirates, 
we are obliged to imagine that from original aspirates they 
became tenues in a-'x^, a-0 ; from tenues, spirants ; and then 
from spirants, tenues again. Because if x and were never 
spirants in the groups a^ and a-0, how can we account for the 
existence of a spirant ^ in acpl For if the popular ar, ax of 
Modem Greek in place of a0, a^ is to be regarded as a survival 
from Ancient Greek when a6, a^ became err, o-« through loss of 
the breathing, how is it that we have not also a modern air to 
correspond to the ancient acj) = sp + h? or, in other words, why 
of the three ancient aspirates has (f> alone become a spirant after 
a, whereas and x after a are supposed to have lost the breath- 
ing in very early times and to have remained pure tenues in 
this combination ever since ? 

That which has become a definite and fairly universal law 
in Modern Greek for the pronunciation of x and 6 in ax, o"^, 


and which will in course of time presumably also become so for 
a-cj), was only in embryo in ancient times and not firmly fixed 
in the middle ages. In Prodromos err or ctk seems difficult to 
find, and in the Chansons Populaires (Legrand i.) from the 
MS. de Vienne of the fifteenth century or earlier we find o-t, <tk 
or a-O, a-x written indifferently, e.g. aa-deva and SpocrtcrTrj (p. 34), 
a-Ko\ij, opyia-ry (p. 48), probably i^ovpia-Tm (p. 62) da-diveia 
(p. 68) from the fifteenth century, ayfuia from the twelfth 
century (?) ; [and from the beginning of this century c. 1820 a.d., 
we have eTndcrOrjKav and icarrjpda-dfjva (p. 146), ^tocrOfjTe p. 150], 
and from Wagner's Medieval Greek Texts prior to 1500 a.d., 
dtrdeveia passim, <T')(io'Sriv, St^ao-^jjz/, dcrxokriaiv, axpiviv, 
KaXv^itTTovv, iyekda-drjv. In the tenth century we get the 
form (7j^ajo^«H' = ets xApiv, and fioaxp^oXave and ^etr^t^ovra? 
from the seventeenth century, and from Koraes, "Atuktu ii., 
we get the following words which are anterior to the fifteenth 
century : a-Ka^m, o-kw, a-Kapi, = 6a-j(,cipiov, a-tcaroyepoi = 
e(7')(aT6yepo<; and e0avfida-Tr]v. We notice that aK is found 
more frequently in the middle ages than or and, conformably 
to this advance on the part of ctk duririg those ages, we find 
that it is the only group of which can be said nowadays that it 
is always spoken instead of a-x- The transition of a-0 into err is 
not yet complete; thus in the beginning of this century we 
find in Legrand i. i-mda-OrjKav v. Karrjpda-Orjva p. 146 and 
^(oaOrjTe p. 150, the word irpoaOev is seldom pronounced 
•wpocTTev, and in Michalopoulos' 'Kafiara, a collection of modern 
popular songs, we have ^iXiovfiaaOe, Kot/iacrOe, ^laa-drJTe 
intermingled with similar forms written with the tenuis after <7, 
i.e. err. This evolution of tenues out of x'P^ after an im- 
mediately preceding a has then, as we hope we have not 
unsuccessfully attempted to show, been in progress ever since 

F 2 


the fifth or sixth century B.C. ; between that time and the 
present we cannot lay our finger on any period and say " At 
this period there is no instance of (tk, <tt, o-tt for o-j^, aO, <r<f>, 
but X ^ were evidently being spoken and written as spirants 
in these cases, and it is only from this time onward that the 
spirants after a gradually changed back into tenues, a transition 
which has not yet reached completion." No, we certainly 
cannot do this, and hence we seem obliged to conclude that 
X<f>& must have been spirants before they ever began to be 
changed into tenues after o-. Take, for example, the word 
a-Oevoi — say this in Attica in 400 B.C. was properly pronounced 
st + henos, in 1890 it is pronounced " stenos " or sometimes 
" s-th-enos," and soon after 400 B.C. we find it is already some- 
times pronounced " scenes " there as shown by the wrongly-spelt 
inscriptions 'AXKia-revov, KaiacrTivov, '^ina-Tevov, yet in the 
middle ages when 6 was confessedly a spirant we have acrdevd 
and aaOeveia, (i)^aa6ev&. Now when and how did spirant 6 
get in ? Did it already exist before 400 B.C., and was (rBivoi = 
"s-th-enos" then and not = "st + henos," or were two changes 
going on simultaneously in the same word ? That is, was o-^^i/o? 
on the one hand becoming o-rei/os without aspiration, and on 
the other hand gradually changing its ff from an aspirate into a • 
spirant, so that by 1200 A.D., let us say, the two forms, a-Tivo<; 
and a-Oivoi with spirantic 0, existed side by side ? That such 
contrary tendencies should have been at work at the same time 
in the same dialect is highly improbable, and therefore if we do 
not wish to assume that x<f>0 were spirants even in the fifth 
and fourth centuries B.C., the only other explanation of these 
double forms which have stood side by side from that remote 
period down to the present day is that of dialectical preferences, 
in other words, that in some dialects, e.g. the Elean and Jioprian, 


the aspirate after o- was very early replaced by its corresponding 
tenuis, whereas in others, e.g. in Attic, the aspirate held its own 
in this position until in the course of ages it here as elsewhere 
gradually developed into a spirant. Then after Christ when 
the Koivi) SiaXeKToi}, a mixture of all the dialects without the 
distinctive features of any one, was the Greek commonly spoken, 
these dual forms from the various dialects were also gathered 
in and became common property, and were then used indifferently 
for some time until, as the modern language shows, the forms 
in t7T, a-K ousted those in <t9 and crx, while the sturdier a-<f) 
still remains, though that too will sooner or later share the 
fate of ad and o-j^. 

If this second explanation be adopted, then the examples we 
have from Attica of <tt, otk for ad, a-^ must be regarded as the 
work of strangers who spelt according to their own dialect. 
Altogether though, this explanation does not seem satisfactory, 
for, seeing the many different dialects which give us forms in 
a-T and o-k, it seems strange that the corresponding ones in ad 
and ax should have had the force to survive right into the 
middle ages and even longer, and also that air for a<f) should 
not be universal in the modern language as aK and ar are. 
For here we must draw attention to the fact which we only 
just stated before that acf) is still spoken throughout Greece, as 
is admitted by Foy and even the great popularist Psichari. It 
has only in Trebizond become air in some words, and vice versa 
IT has in a few words been changed into ^ after a, e.g. the 
Naxians and Gaeopontines say a^vpi(; for ordinary Greek 
aiTvpi'i (both these forms are old). We also find several 
varieties, a^ovTvki, airovhvKo'; and &(ji6pSvKo9, and the popular 
forms 0-^0771' and a(f>oyydpt (answering to the ancient 
airoyyiov and awoyydpiov) as well as the old form 0-^0770? 


which is known still to exist in the present spoken language. 
But except in these dual forms, cttt is not spoken for a-<j). 

Thus this non-existence of o-tt for acf) (except in Trebizond), 
parallel to the <tt and o-k for ad and ax, militates most strongly 
against the supposition that in the modern ar and ax we 
have the survival of the explosive component of the ancient 

All things considered, it seems as if we had in this case a 
chain of evidence to show that from very early times x'f'^ were 
spirants, and that a preceding a caused their change into the 
corresponding tenuis, as it still does ; or even, if this is too rash 
a deduction, it is at any rate clearly evident that in many cases 
the % ^ immediately following a was from very early times 
changed into the tenuis, and that consequently this change 
which for two of the three sounds regularly takes place in the 
modern language cannot with any justice be styled "a new 

Another difference in phonetic law is said to be illustrated 
by a surd spirant not allowing a preceding nasal, so that in the 
language as now spoken we get such forms as a60o'; for av6o<;, 
vvif>7) for vvnjtrj. 

But if we look a little more closely into this, we shall see that 
this phenomenon in the contemporary language, far from being 
a new phonetic law, is likewise but an extension of a modifica- 
tion that had already attained considerable currency in Ancient 
Greek, to judge only from such examples as have survived. 

But there is another reason which argues still more con- 
clusively against calling this a new phonetic law of the present 
language, and this is the fact that the modification we are now 
to consider is only part of a larger one, begun in ancient times 
and still in force, according to which nasals were omitted, not 


only before aspirates, but also before mediae and tenues, and 
that from the sixth century B.C. downwards. 

Blass, in speaking of this omission of the nasal in Ancient 
Greek, remarks in a footnote (p. 87) that " this rejection of the 
nasal appears in Modern Greek too, but only before x4'^ owing 
to a special tendency." We venture to think that this remark 
would give a wrong impression to any one who did not know 
Modern Greek ; he would imagine it meant that the nasal only 
disappeared before %^^, whereas it constantly does so before 
tenues and medials also, just as it used to do in Ancient Greek. 
Examples of the omission of the nasal before a tenuis in 
Ancient Greek are Tviravov, poetical form for rv^iiravov, Eur. 
H.F. 888 &c., evKa/j.ire'i which is scanned as a dactyl in Anthol. 
p. 6, 4 ; so also aymKaKr\y,wTa scanned as atrXaKrifiara in Eum. 
934, and avaixirkdK'qTo<i which the metre requires to be read as 
avaTrXa.KrjTO'i in O.T. 472. And from inscriptions 'OXuirto?, 'OXv- 
'jri6Scopo<; &c. passim in C.I.G., 'EKe\aBo<i on a vase C.I.G. 8182, 
and 'ATdXaTTj ib. 8185 on one of sixth or fifth century B.C., 
TvTcipeoi ib. 8220, Oypr. raXaTtov, arl cf. Coll. i. 60. Corre- 
sponding to this we find in Modern Greek oTiva = ovTiva, koto^ 
= Kovrof, Kavco = Kafivca. Psichari is not quite exact in saying 
" V est rest^ partout oii il ^tait devant une explosive ancienne 
e.g. irdvTa." 

Before mediae it is also omitted (or assimilated) in both 
Ancient and Modern Greek, e.g. ^v^^aXXevdat C.I.A. ii, 52c, 
Pamph. aSpt, yevoSac, -TreBe (^Trevre) &c. I.C.A. 505, Delph, 
"A^a/S/So?, Ti/idBpa (=:TifidvBpa) Meisterhans § 31, Ka/3^d(} a 
var. lect. for ica/i^d^ = Kara^d<i in Find. N. vi. 58, and Kv^^a a 
by-form of Kvfi^ri. 

In Modern Greek fidSpa = fidvBpa, Kv^aXa = KV/i^aXa, 
Lastly, the omission of the nasal, especially v, is very frequent 


before 'x^ij) 6 in the present language, e.g. vv^rj (= vvfi(j)r}), a6o<s 
(= avOo'i), avvwpSs (= avy^copfo), KoKoKvdi = koKokvvOiov, 
fj.a6dvca = /Mavddvco and many others. But it is by no means 
regularly observed in every case and, to speak impartially, one 
might almost say that the forms without the nasal are not much 
more frequently heard than those with the nasal, and some words 
there are in which the nasal is never omitted, such are ^av06<; 
and its compounds as ^avOovXa. dvOo'i and dv6r) are quite as 
common as dOOo';, aOdrj, avOl^m more common than dOi^ca, 
dvdpa-iro^ and dffpawo'; used indifferently, ivdvfiaa-ai rarely, if 
ever, becomes iOvfida-ai, vutjirj always spoken for vvfi^rj, and so 
on. In the extant poems of the middle ages there are very few 
words to be found containing a surd spirant preceded by a nasal, 
in Legrand vol. i. ■jravddvo) and dOovcn occur in a poem prior 
to the fifteenth century. 

This omission of the nasal before % ^ ^ can be paralleled by 
similar spellings from Ancient Greek which are found in most 
varied places and times. Like the modern vv^ we find on 
vases of the sixth and fifth century B.C. d(f)i (— dficjji), vv^r]<; 
and vv^ai {cf. Meisterhans § 31), and on an archaic inscription 
from Siphnos vv<f>e(ov I.G.A. 399 ; also Nu^oSw/ao? G.I.G. 3155, 8 
and 'A<}iiepeco<} G.I.G. 7710 ; and on the Corinthian clay tablets 
(Rohl, D.I. 3119f) Afi(j)tTpiTa is twice written with an /m, twice 
with V, and twice without a nasal, which shows how weak and 
uncertain the sound of the fi must have been — also 'A^iapffo'i 
ib. 3140. 

Like modem av^mpSy we have on an Ionic papyrus of uncer- 
tain date, probably of the time of the first Punic War (c/. 
Petrettini, Twp. Greoo-Egizj. 1. 15), Tv^dvoi and TV')(xdvoi for 

In Cyprian the nasal is never written before a consonant and 


so this dialect has aOptono^ (= dvdpcoTros:) and Tpe/M6ov<s which 
probably stands for Tepfj.ivdoO'} ; these can be compared with the 
modern adpanro<;. Like modern a6o<i, iiaOaivm &c. are Ad-n-vOog 
from Corinth (Rohl, D.I. 3132), and S/ii'^to? ('ETriyovov) for 
lifiivdioi; on a Rhodian amphora found in Athens (cf. Dumont, 
Inscr. dram. p. 92) and again QAkvmv) 'tp>lBLo<i on another 
Rhodian amphora from Olbia and dating from the third or 
fourth century B.C. ; also 't/itdcov (= 't/J-ivOcov) on a Melian 
inscription I.G.A. 413. Further Stephanus Byz. gives us the 
two forms aiyiOoi; and a'iIyiv6o<; : the former is found in Arist. 
II.A. ix. 1, and Callim./>. 321, and the latter in Oppian Ixent. 
i. 10 ; iii. 11 and is suggested as an emendation in Antig. 
Oaryst. ii. 

As further examples of the undefined pronunciation of the 
nasal we may add iS^t^ for %^ly^ G.I.G. 8139, K^oaTavTlvoi 
G.I.G. 9025 and Kwo-raz'Tto? by the side of the more usual 
form K.a)va-Tavrlvo<; ; in this last word this omission of v is due 
to the Latin pronunciation of Costantius, Costantini &c. {v. 
Seelmann, Aiissprache d. Lat. pp. 283-4). In Modern Greek it 
is also dropped in some dialects between vowels, as Lokr. 
6«eto9 = e/eetvos, Kaevaf = xaviva^ ; at the beginning of words, 
'A^la=Nd^o<;, "EiraxTO^ = NauTraxTo? and constantly at the end 
of words as tt] Kopr} = ttjv Koprjv. 

It appears then that the nasals still retain the tendency to 
disappear before a following consonant which they had in the 
ancient language— it is not clear to us why Blass says their 
omission before j^ ^ in Modern Greek is due to a special 
tendency because, as the examples we have collected show, they 
are omitted before other consonants as well and always have 
been more or less so from almost the earliest historical times. 
Hence since the nasal is omitted in Ancient and Modem Greek 


alike before tenues, mediae and % ^ ^, though the examples of its 
omission in Ancient Greek that we can collect from inscriptions, 
vases, &c., must necessarily be very few compared with the 
numbers in Modern Greek to be gathered from the lips of any 
uneducated living Greek, we hold that its omission before % </> ^ 
in the popular language of to-day is not a new phonetic 
departure and cannot therefore be used as evidence of a change 
having taken place in the pronunciation of ;)^ ^ since the 
classical period. On the contrary, it seems rather to point to 
the similarity of the pronunciation of % ^ in classical and 
modern times. Psichari goes so far as to say that " if the date 
of the first instance of the omission of v before ')(^<f> or 6 can be 
ascertained, that will be the date of the transition oi -^^6 from 
aspirates to spirants," so that according to him the vvcl}r) or d(j)i 
on Attic vases of sixth or fifth century B.C., the rvy^avot on the 
papyrus of Artemisium, or the %fji,idio<; on the Khodian vases 
would conclusively prove that in these dialects at least x4'^ 
had already become spirants. 

However, without going so far as that, it may at any rate 
be justly maintained that, even if on the one hand the omission 
of the nasal before % ^ ^ which prevails both in Ancient and 
Modern Greek is of no value to prove the spirantic nature of 
X(f>d in ancient times, it cannot on the other hand be used to 
prove the aspiratic nature of these letters for ancient, and their 
spirantic for modern times. One and the same fact cannot be 
taken to prove two directly opposite things. Perhaps this fact 
cannot help us at all, because as the nasal is omitted in Ancient 
Greek before a pure tenuis, e.g. the /* in rvtravov, it could 
probably be omitted before % ^ ^ just as easily if they were 
" tenues + spiritus asper." 

Lastly, a new phonetic law is said to be responsible for 


tbe fact that a j^, ^ or after p is sometimes changed into a 
tenuis in the modern language. 

This change is, however, at present still the exception and 
by no means the rule, as it is only in very few words and 
then not in the ordinary Greek that this change ever occurs : 
Psichari (c/. Mim. d. I. Soc. Zing. Paris, vi. p. 304 ff.) says 
distinctly that forms like oprcovco and epKOfiai are not common 
in Greece and not the rule even in Constantinople. Even ^pra 
and epKOfiat,, which are perhaps the two such forms most 
frequently heard, are in continental Greece less common than 
fjpda and epxo/^ai. In fact it would be more correct to speak 
of this change after p as an exception to the general persistence 
of p(j>, p^ and p6, for in the majority of words they remain 
unchanged, e.g. in Kapt^iraa, irapdiva (demotic for irdpdevos;), 
apdovvia, ap')(iv(o, &c. Further too, the \ before ^Q is in 
popular Greek generally changed into p, which shows that the 
combination p^, pd is by no means disliked ; thus aSe\<j)6<; 
becomes ahep^o'; (never aSe/STro?), Be\(f>iv Sipcftiv, and ^X6a, 
ek6r]<i &c. become ?ipda, ep6ri<; &c. ; also Kop^o^, which is popular 
for K6X'n-o<i, and Kopcf)^ contraction of Kopv^rj. The change of 
\ into p before 6 and ^ had begun in the middle ages, as in the 
poems before the fifteenth century we find ?ip6a, ep0rj<;, &c., 
passim. This change, which is so frequent in mediaeval and 
modern Greek, indeed entirely disproves the existence of a new 
phonetic law, according to which a spirant after p becomes a 
tenuis, as it proves that so far from the sounds pcf), pj^, pd being 
disliked, they are on the contrary rather favourite com- 

We d& not think, therefore, that any new phonetic laws 
have come into existence, but that the beginnings of laws 


already visible in the ancient language have been developed 
so considerably that at first sight they appear to have changed 
the character of the language. 



We will now see whether, allowing for natural modifications 
due to the lapse of centuries, the sanae interchanges of sound 
still occur as occurred in ancient times. 

Beginning with the interchange of aspirates with one another, 
we find {a) a frequent substitution of ^ for 6 in ancient Greek 
in the Aeolic dialect. Thus we have (f)iX6(f>eipo<; (Inscr. inidit. 
Ussing No. 25) from Larissa, which, according to Meister, is for 
^LXodrjpoi;, while Ussing himself says it is for <f)iXo')(eipo<;. On 
a Boeotian inscription of the fifth century B.C. we have <E>eTaXos 
(Meister, Gr. Dial. i. p. 204—211, Tan. 49); also ^er-rako^ 
{ib. Theh. 28, 8) on one of the fourth century B.C. both for 
&eTTaX6<; (G.I.G. 2430), while the regular ©erraXov occurs 
(ih. Thesp. 27, 3). Similarly &€6^ea-To<; and %i6j>ea-T0<s (CJ.G. 
3172, A. 42, B. 91) of the third century B.C., which Blass 
interprets as = @e6dea-To<; (perhaps due to dissimilation). 

We also have the Boeot. and Epir. (j>ea)v, and ^vov7e<; for 
6e5)v, 6vovTe<i (v. Cauer^, pp. 174 — 5) ; and this spirantic sound 
of in Boeot. 6e6<; Aristophanes {Ach. 905) has indicated by 
the un-Boeotian writing crto? (c/. Meister i. p. 260). Alcman 
has the Aeolicisms ^oivai^ ( = 6oLvai<s) fr. 24, B.^ ■n-okv^oivo<; 
( = ■KokvOoivo'i) fr. 14 Bergk. 


Homer, Pindar, Theocritus, Aristophanes, &c., have ^Tuiv for 
6\av, also (^Xl^eTai for Bki^eTai Homer and Theocr. xv. 76 ; 
these forms Curtius regards as ordinary Greek. Meister says 
that the Homeric and Pindaric ^j;p and <prjplov are not to he 
identified with drip and 6r)piov, they do not occur on Aeolic in- 
scriptions. The grammarians reckoned these forms as Aeolic 
because they considered the old population of Thessaly was Aeolic, 
£|,nd ^jfjo in Homer is only used as the name of a Thessalian 
tribe, and in Pindar of the Centaurs. On an old inscription 
from Naxos, we find Aw/ao^ea for Acopo6ea (I.G.A. 411). 

In modern Greek the substitution of <f) for is found in a 
few words when followed by a vowel, X or // (the last in these 
cases being changed to fi) — thus from Trebizond: ^Xi^epo<; 
( = 6XiPepo<i), ^aXafiih ( = 6a\afiiSiov) ; and in the ordipary 
language api^vrjroi ( = avapiO/irfTOs;), a-Ta^vt] ( = a-rdd/iri), 
(prjKapi ( = 6r}Kdpt,ov) and @fj^ai is called ^rj^a by the people 
of that district, (b) There are very few instances in ancient. 
Greek of the substitution of for ^, and according to Meister no 
inscriptions give us such a substitution. Hesychius gives iXaffpd 
(= i\a(f)pa) ; also 60pvv which, he says, the Cretans use as meaning 
a mountain, and which is probably a dialectical form of the 
original o'^jou?, and there is the Cretan 0vX\.a, which may be 
for ^vXKa, though Curtius says it is doubtful, and more doubtful 
is the identity of the and ^ in Kopv0- and Kopv<^ri respectively. 

Similarly, in the modern language the substitution of for ^ 
is seldom found. Thus there is ffXixr], a form for (fyvXiKt], This 
change is rare, however, except in the Tsakonian dialect, where 
it is frequent before vowels, p and \ : thus dXoi0^ = aXoKJ}!], 
vv0r] = vv/i(l>i], evfiopdia = eufiopt^la, and generally in this 
dialect 0i,, 0v = <jii, (j>v respectively ; thus, 0iXe = ipcXois, dvre = 
^vToVy ov0e, = ofi'j. But, as Carl Foy says, no one would here wish 


to recognize an Aeolian survival. Besides this dialect is very 
peculiar and distinctive from all other present Greek dialects. 

This interchange of 6 and tj) can only be explained by a 
spirantic pronunciation of the two letters. Thus Roscher draws 
our attention to the substitution in England by children of 
/ for th, thus iumi instead of thumh. There is another word, 
which, if we can consider the forms that occur in dififerent 
editions authoritative, goes far to prove the spirantic pro- 
nunciation of and <f). This is the word Fibvi (vid. Hermes 
xxvii. p. 481) (thus written on an ancient Corinthian vase) 
which occurs in Aristotle's Naf. Hist. i. 18, p. 617, 1. 9, and is 
written ^cov^ in some MSS., whilst in others a form dSiv^ is 
given (yid. Bekker's edition). Thus both 6 and (/> here stand 
for the original sound f or hav. 

We see, then, that the interchange of 6 and </> as exemplified 
in the ancient dialects and the modem language correspond 
very closely, and we have no reason to think that the cause 
of such an interchange was not the same in both cases, or, in 
other words, to suppose that the pronunciation of sounds treated 
in the same manner at both periods of the language was 
radically different at the one time from what it is at the other. 

(a) 6 became ■x^ in ancient Greek in the Doric dialect, thus we 
have : ef ej^a = e^adev, i^ev'^m = i^iXBco, ixH-ct = Wfia {vid. Hesy- 
chius). i-^^fiara, idfiara and t^j^i'ta are var. lects. in II. xiii. 71, 
and opvi'x^o'i &c. for opvido'i &c. in Pindar and Theocr. We 
have also the Lesbian irX'^'x^co = trXijOco (vid. Cramer, Anecdot. 
Oxon. i. 149, 6). 

Corresponding to this we have in the present Cappadociau 
dialect -^ substituted for d before and between vowels. [It will 
be seen that several of the above examples were instances of 
the d between vowels being changed to ^.j Thus, for instance ; 


^€0? = 9e6<i, %eXo), ')(aKaaaa = 0iKco, daXacraa, "yavaTrnva = 
davaroeo, ya^Kco = ^ay8%«i) (vulg.) =s Bdirrm ; and ^d^o(}, crTrj'Xpi 
= 0d6o^, a-Ti]do<;, ftep^w = neOvm, Kpiyapiv = KXiOdpiv s= KptOi], 

It is also interesting to notice that in some MSS. of tlie 
Septuagint dpovov is found for ^povov in Ps. 8, 46 and ^ripav 
for dripav in Ps. 131, 15. 

Neither in the ancient nor the modern language do there 
seem to be any instances of (&) 'x^ being changed into 6. 

Of (a) (j} being changed into ;i^ we have from the ancient 
language the word dpxiSav'xya(^opeiaa<i found on some 
Thessalian inscriptions for dp'^^iBacpvrj^op'^a-a'}, in which Savxva 
consequently stands for Sd<f>vr] ; of these two forms Curtius 
opines that Sav'x^va the Aeolic form was probably the earlier. 

Hesychius gives a form Savx/J>6v (Hipponact. fr. 2) which he 
says is for Bd<j>pivov ; Meineke here reads iravSav'xycorov and 
Bergk roiovSe hd<^vr]<s. Alcman also has Bavj(v6cj}opov for 
Ba(f>vi](fiopov fr. 19, B. 

In modern Greek we have 'XTivoirmpo for (^dtvoiTtopo, and 
Lesbian d^avo^Trji} for d^av6<f)Tq<;, which is the same word as 
the ancient (^avovTrfi which occur in the Scholiast's explanation 
of Aristoph. Eg^. 997. 

(6) j^ becomes ^ in ancient Greek in the Lesbian avifyqv =: 
avxnv, of- Meister i. 120. In Theocr. xxx. 28 {vide Paley's 
edition) Fritzsche reads afi^rjv, another Aeolic form, which 
Salmasius restores to av')(riv. We have irvpi^ov = Trvpiy^^ov 
C.I.A. ii. 2609 in Attic. Hesychius mentions the form <f}Xiap6<i 
a by-form of ■)(\i,ap6<; in Herod., and this last word gives us a 
double indication of a spirantic pronunciation, if it can be 
rightly connected with the form Xiap6<i (vide Liddell and Scott). 
Prellwitz, however, does not admit this derivation but takes 
Xta/309 to be derived from a different root, Vslaivo. Hesychius 


also gives Ka<f>d^eiv and ku'x^, Ka>(f>eveiv and «wj^-, in which, 
according to Curtius, ^ is the earlier. 

In the contemporary language we find ^ changed to ^ after 
7\, thus 'kelya = yXeicfxo ; ^\ri')(mv and r^Xriymv = '^XTj^wvaKi ; 
whilst in Calabrian ')(t ( = kt) becomes ^t, thus : vv<fiTa, 6<f>Tm, 
avoi^To for vvyra ( = vvkto) etc. ; and in the same dialect ^z/ = ^v. 

We see then that the mutual interchange of aspirates is not 
frequent, either in Ancient or Modern Greek, but, such as it is, 
the mere fact of its existence in several different dialects (e.g. 
Aeolic, Doric, Cretan) proves that the aspirates interchanged in 
the respective dialects had in these a spirantic pronunciation. 

To pass on to the interchange of Aspirates and Mediae, 
(a) Beginning with S for 0, we find that in the ancient language 
the Macedonian hdvov = 6dvaTo<; ; and cfuSaKvoov is for iridaKviov 
on an Attic inscription of 330 B.C. (G.I.A. ii. 807, b 114, 117). 

In the modern language this change is comparatively frequent 
in dialects. Thus we have the Cyprian aSpmiroi; = dvOpeoTro's ; 
dvaSpfJKa = vdpOrj^ ; the Locrian Aio'^dprji;, At60t\os = 9eo')(dpyi<i 
and de6^iXo<s, also Bv^arepa, which occurs in other districts as 
well. On Leukas, Thera, etc., we find the forms MdpSa for 
MdpOa, A6/JL0KO for &avfj,dKia (a town in Thessaly) and Seid<piov 
for 0eid<j)iov (Passow, Carmina). 

(h) We find several instances of 6 for S in ancient Greek. 
Thus, 6da-o<; = hd(TO'i (Steph. Thes.); in the word @v<})aLOiSri<; 
(of. Meisterhans § 39) we have an Attic form dating from the 
sixth century B.C. of T!v^aiSiSr)<;. On another inscription of 
373 B.C. ov6' ol = ovh' 01 in Attica (c/. ib.) ; and probably 
50' 'Ep/ifji = SB' •Epfiv'! (G.I.G. i. 12), also ov0h = oiBiv ; 
similarly in New Attic forms we frequently have for B 
in ov0ei'i, fir/dip, &c., for which the feminine is always ovSefiia, 
firfBejjbia, never ovTefila, iirjTefjiia, but all these cannot perhaps 


be taken as true examples of interchange, as the 6 is here due 
to the folio-wing " spiritus asper," v. Brugmann, Greek Grammar, 
p. 52. It is worth noticing also that in some MSS. of the 
New Testament we find i^ovOevSt ( = i^ovBevS)), of. St. Luke 
xxiii. 11 and Eom. xiv. 10. 

There do not seem to be any instances of 6 for S in the 
contemporary language. 

In the ancient language (a) <j) became /3 in the Macedonian dia- 
lects, thus we get Bt'XtTj-Tro? ( = Ot'A.tTr'Tro?), cf. Etym. Magn. 179 : 
^epevLKf], Bjouyev, Ke^aXr/, d^povTe<; ( = 6^pv<s), l3a\aKp6<; (Diod. 
Sic. xvii. 55). In the modern language there is dXei^co for dXei<f>(o 
and the Cyprian /SXd/io? ( = (p^6fji,o<;) and ^puKTrj ( = (^paKTrf). 

In the ancient language the only instance of (6) ^ becoming 
<f} appears to be in the word BaXtd? (the name of Achilles' horse 
in Homer) = " bright," " gleaming,'' of which a by-form ^aXio^, 
explained by Hesychius as being the same as BaXtd?, is found 
on ancient vases {e.g. 0. Benndorf, Wien. Vorlege. Matt. 1888, 
Taf. vi. 3a) and (Gerhard. Etrush. Vasenbild. xii. p. 17), and 
this form is also found in Callim. /r. 176, compare also Thuc. 
i. 24, and (f>d\,ap6<; in Theocr. v. 103, viii. 27. Prellwitz does 
not connect these words. In the modern language the only 
instance of this interchange seems to be <^Xrj<TKovvi = ^Xri-)(pvvi 
= ^Xri')(a)v, of which we have the Ionic form yXtjx^^v in 
Theocr. v. 56. 

In ancient Greek (a) we have no instances of % becoming 7. 
In the modern language only sporadically in dialects ■^ 
becomes 7, thus rypova-6<; = ;\;/3i'o-d?, Cypriot ypova-dcf>iv, yp6vo<i 
= vpovo^. In Tsakonian this change takes place between 
vowels, e.g. dveyov = dve')((o. 

In ancient Greek (b) 7 becomes ;>(;, e.g. in irpfixt^a for irpdyn.a 
{I.G.A. 3816 17) fifth century B.C. on an inscription Cr(?pi(^l>ios;_ 


also from Attica of first century A.D. (?) •n-p'^xf^aro's {G.I.A. iii. 
3822, 4); irapdhefxjJ.a (Ephem. Arch. 1886, p. 166, 1. 251) 
second century B.C. 

In the modern language y becomes x only in Cypriot irafx^viSi, 
and the Locrian Sv^^aripa for Ovyarepa. 

We see then that there are but few instances of interchange 
in either the ancient or the modern stage of the language 
between the aspirates and the mediae. 

We next come to the interchange between the aspirates and 
the tenues. (a) For d becoming t in ancient Greek. Hesy chins 
gives the change before p, thus rpova, for which we have Opova 
( = flowers) in II. 22. 441, only given by Hesych., Tpvyovav, = to 
tap at the door, a var. lect. for Opvyovav Ar. Eccl. 34, and rpiva^, 
a by-form of dpiva^ which is found in Anth. vi. 104, 6, 
cf. also Hom. @pi,vaKir) and later TpivuKpia. 

In Elean we have evravra for ivravOa, whilst the Ionic 
form is ivdavra. Old and New Ionic give us aSri? and a?j6i^, 
of which two forms, according to Curtius, the priority cannot 
be determined. 

Athenaeus has Hvrvia for Kvdvia. There is besides the 
regular change of to t after a in Locrian, and sporadically 
in Boeotian and Phokian, also in Elean, thus : rifj.a)aTa>v 
(Tl/ji,da-0ci)v), KeXoiarav {-adrfv), XvadcrTO) {-aGw) from Elis 
[Roehl, I.a.A. 117, 119, 121]: ^meo-TO), eXeo-rat [I.O.A. 
Oeanthea 822], the Attic Kidapi^eaTco [G.I.A. iii. 74, 3], 
diroiroKnevaacrrai — irapayeveaaTat Bull, de Gorr. hell. v. 
p. 398, the Attic forms 'EiiriaTevov and AaKia-revov, and the 
Aeolic eo-TO<?, fiaa-ro';, klctto^, Klarapo';. 

From Attic inscriptions we may give as instances of this 
change evrvfiia G.I.G. 708; on a vase BaruXXo? {ib. 8439) 
KaTia-TOLo-iv Rang. i. p. 62 and from monuments of other 


dialects toi/ reov ( = rov deov) G.I.G. 3993 ; evrdhe (ib. 1988c) ; 
'Ardfia'} = 'A0dfia<s on a Mysian coiHj and many others. 

In the modern language we find Xevrepo^ for eXevdepoii in 
the Peloponnese, reXeo for diXm in Asia Minor. Between 
vowels, corresponding to the ancient Ionic avTus, we get the 
form euTw? for evdv';. Again like in ancient Greek, d is changed 
to T after tr, e.^r. ytito-To?, aiardvofiat,, darev^i. The change is 
frequent after ^ and pj;, giving the combinations ipr and ;;^;t ; 
of the latter we have already (p. 57) given one probable 
instance from ancient Greek ; it is comparatively rare after p, 
e.g. ?ipTa = ripOa = ■^\0a ; oprmvo} = opOcovco. 

(&) T becomes in ancient Greek, in Boeotian which has 
-vdi = -pTt = -cTi for the termination of 3rd pers. plur., thus 
eycovdi, dirohehoavdi ( = diroSeSdoKaa-i). 

From Attic vases and inscriptions, Kapidaio<i = Xapnaiof 
K. V. 51 ; ®v(patSiSri<; = Tv^aiSiSrj'i K. V. 97 of sixth century 
B.C. ; yiOaiv, rnddav — 'x^itoov of second century B.C. ; Opo^o'; = 
T/3o</)o? G.I.G. 8139 ; 'ApiaTOKpadei; on a vase {Lpsg. Schrft. 
viii. p. 747) and 'AvdCKo-xpi {K.V. 51) of sixth or fifth century 
B.C. ; and others such as old Attic ivOavdol and dveOidri (G.I.A. 
iv. 6 27), 9efiL<T6oKXri<; {ib. ii. 864) of fourth century B.C. 

From Cumae we have 9v^X6<s (I.G.A. 624) and i6edr)v {ib. 
528) ; from other parts there are several cases of 6 written for 
T, e.g. 0evSo0o-; = 0evSorov {G.I.G. 8518). 

Plato {Grai. 406 A) gives the form Ari0a), as one used by 
^evoi for Ar/TO). In Ionic we have ^d0paKo<! for ^drpa')(o<;. 
We may also notice double forms as irXdravoi} and ■7rXd0avo^ ; 
and the terminations -rpop and -0pov, -rpa and -0pa, -tXo- and 
-0Xo-, -tXt) and -0Xr], which correspond to similar Idg. ones {v. 
Brugmann, Gmnd. ii. 115). In the modern language medial r 
regularly becomes before vowels, e.g. ■)(apdiou ^-. -xapTiov ; a^so, 

a 2 


IXiovde = fi^re in the Cretan dialect. In Oinoe £k jeve6r)(} for 
e'/c yeveTri<; ; and in Tsakonian a-i]da = afjTa. Other instances 
are : cWi^okij = avTi^oKrj, fjueOavptov, aTpaOicoTr]'; and jtadia, 
which are sometimes used in the vernacular for the forms with 
T ; and 6pe<^(o and 6po(^ri are still said, and correspond exactly 
to the ancient dp6(f>o^ for Tp6(j}0<;. 

Passing on to the labial we find that (a) becomes v in ancient 
Greek in several instances apart from the combinations a^, air. 
Thus afiireaai is Laconian for a/x,(pieaai, ; in Doric Trdrvrj = 
(^oLTVif], also dfi'TTidovpo^ and dfiTrierTaTrip. 

The form dfiire'x^ei.v with its derivatives belongs to all 
dialects where the ir is due to dissimilation. There are also 
the double forms pdirvt; and pd^vi Curt. p. 502 and Tj-pooifiiov 
= (fypoifMov, the latter form prevailing since Aeschylus. 

There are hardly any instances of tt for (j> from inscriptions 
on vases. In Attic niXnro^, AtTrtXo?, Ni/coTrtXe (G.I.G. 8076) 
are found on the same vase ; ^virpoavvr} (cf. Rosch.) ; Tiava- 
Kkeov; Rang. 1823 ; Mo7ro-09 on a vase. 

Corresponding to the ancient double forms, we have in the 
modern language pairdvi and peirdvi for pd(f>avo<;. There do 
not seem to be many other instances of the change of <^ 
into TT, but we get TrdirXaifia for i(f>d'7T\co/j,a, and dcTTaTrt'Se? for 

(b) The change of tt into cf> is likewise not at all frequent. 
In Attic of the 4th and 5th century B.C. ^apdevo<} {G.I.A. iv. h. 
373) ; ^epae(j)6vr] and Aio(f>6idrj'; (ib. ii. 835 c.) from an inscrip- 
tion of 320 B.C. 

^iTTaKoi; occurs for IltTTaKo? on a Mitylenian coin (cf. 
Mionnet, Suppl. vi. 64 no. 82). There is also the Arcadian (cf. 
Coll. 1181) and Delphic 6e\(l>ov(7io^ (W. and F. Inscr. de Belph. 
464, 5) which on coins is OeXTrovano';, 


In Doric and other dialects e<piopK€Q} occurs for iiriopKeo), in 
Locrian ^piv = irpip, 7/047709, probably the original form,= 
7/34^0? (Curt. p. 354), and the Attic ^tSuKvy = TndaKvr}. 
Strabo vii. 315 gives lidpo'i as the older form of <I>apo9 ; and 
the Attic ^ai/o? has also a by-form ■jravo'i, which occurs in Aesch. 
Agam. 280 &c. and is the older according to Phot. Lex. As 
before we must here also note the double forms in o-tt and a-<f), the 
former being more Ionic and the latter predominating in Attic. 

We find this change in the vernacular of the modern 
language, in which such forms as (povxra = irvKTrj, ^eXexovSta 
= TreXeKoiihia, ^Xefip-ovi = TrXevfioviov, K6p(l)0<; = /coX-tto? occur. 
Also in the double forms similar to the ancient, thus a-(j)ovTv\t, 
= airovBvXoii, &c. It is very frequent in the combination ^r = 
TTT, thus ^a(j)To^to, <f)TV(o, (f)Ta)y(^6<; , Xe<^TOKnpvov, &c. There is 
an Epirotic form d<f>iKpd^ofjiai (it has no corresponding ancient 
form) = listen, cf. Foy, p. 31 ; and the Tsakonian dcpoKtovpi 

= dlTOTVpOV. 

There is more frequen tinterchange in both the ancient and 
modern stages between the guttural aspirates and tenues 
than between, the dental and labial. In the ancient Isinguage, 
(a) y(^ becomes k in the Doric ^pv/cr]dfi6<i (Hesych.) ; in Ktriov 
cf. Meisterhans § 38 and Kvrpa, probably only a Siceliot 
solecism, as Epicharmus has %uT/3a (vid. Ahrens ii. p. 82). 

On Doric inscriptions the form BeKOfiai, which is also 
Lesbian and Ionic, and is used by Sappho and frequently by 
Pindar, occurs for the Attic Be-x^ofiao. In Old Attic we have 
pi'^Kw for the Ionic pef^X"^ (Curt. p. 242) ; kv6o<; for ')(y6o<i. 
New Attic also sometimes had « for p^, e.g. fiovKop = iJ'V'Xp<; ; 
Curtius gives apaKo<; a late form for dpa'x,o<s. In New Ionic 
(e.g. Herodotus) oiiKi = ouj^t, 0dOpaKo<; = ^drpa')^o<; ; Kvdpoi 
and KVTpo'i for ')(VTp6<i, 


From errors on inscriptions • and vases we get very many 
instances of k for ^^ from all dialects. A few examples are : 
from Attic, Ev/tetpo?, Kapi6aio<;, kuXkovv, kiOmv and kitkov 
(Meisterhans § 38), from Aeolic Kopayiav, v-n-apKoia-ai^ (O.I.G. 
3524, 40), NeiKofiaKiSa. On the Gortynian inscription we find, 
according to G. Meyer, owing to the a.bsence of a special sign 
for T^, the form Kpifiara = 'y^prifiaTa, e{7nK)opev and avKopev = 
iinjdcopeiv and ava-ywpelv. 

This change of ;^ into k occurred frequently after <t, whence 
we get (TK for o-^, e.g. aKeX,l<; Attic a-'x^eXh, &c., irdaKou for 
-n-daxoi, (Roehl, LG.A. 112) from Elis. 

In the modern language this change is also sporadically found 
in the dialects, thus we find e/co) for exo>, a-roKal^ofiai, both 
these occurring in Rhodes ; whilst in Cypriot we get oKrpo'i for 
e^Opo^, epKOfiai for epj(^ofj,ai, &c. The forms BeKo/Mai and 
KavKoi)/iai occur in the Cretan dialect. 

Again, as in the ancient language, there are the double forms 
in a-^ and aK, both in the medieval and modern language, thus 
cTKcipa = ia-x^dpa, &c. Ptochoprodromos has /xovoKvdpiv for 
jj-ovoxyrpLv, and also the two forms /3ddpa/co<; (iv. 99) and 
^drpaxof (iv. 409), which thus correspond exactly to the ancient 
Ionic and Attic forms. 

We find (6) k becomes ^ i^ the ancient language sometimes 
before t ; and thus we have the forms Taxja ( = raicr^ or Ta-)(Py) 
Elis, Roehl I.G.A. 113, c. 12, Kariyravev Sagalassus in Pisidia, 
G.I.a. 4377, 5, evTayTo<i (Roscher 81, 20); e^ t&v Chios, 
G.I.G. 2241. We have also numerous instances of x written 
for K in other cases. From Attic, for instance, dvrjxoo'i {G.I.G. 
160, 7), XoXt^o? = K6X'x,o'i> ^nd XaxpvXowv = KaxpvXicov on 
vases of sixth and fifth century B.C. (c/. Meisterhans § 38) ; 
also ' AiT'xXairlo'; &c. {I.G.A. 549) and forms like •irdvho'xp'i for 


those with original « as ttwSo/bo?. Athenaeus has XvTvia for 
Kvdvia. The grammarians give arpexv'; as Doric for aTpeKri<i, 
Pindar, Callimachus and Theocritus have the tenuis, though 
one reading for Theocr. ii. 151 is a.Tpexe's and is adopted by 

In the modern language also k becomes ^ before t, e.g. XTeVt 
= KTei<;, dSpdy(^rt = drpaKro^, dvof)(T6<; = dvoticTO'! ; e'p^; Tij^i)? 
in Trebizond. We also get this change before vowels, thus 
'X^d<fjTco = Kdirjm, ■yfri'^aXi^ei, x'>X^''°'' > ^^^ before v, e.g. Sel'x^vm 
= SeiKvvco, Xi.'^vi^Q) = XiKvi^ay. Before tt, e.g. e)(7rdyrjv = 
eKTrdyqv {e^eird'yiqv), d')(Trdv<o = ^Kirrj'-jalva, d')(Trapdaa'(o = 
iKaTrapda-crco in Trebizond only. 

These modern words in which « becomes p(; before v may 
well be compared with parallel forms in ancient Greek where 
the aspiration is considered to be due to the influence of a 
fol lowing liquid or nasal, e.g. Xup^jco? from JXvk. 

liKo'^fio^ „ JifKeK. 
^\7]'x^p6<s „ f/^Xarc. 
From the preceding pages we see that aspirates interchange 
hut seldom with mediae either in the ancient or modern 
language, y and 'x^ hardly ever in either, (p and /3 do so in one 
dialect in each, and and B do so in Attic, occasionally in 
Ancient, and in one or two dialects in Modern Greek — to sum 
up, the interchange of aspirates and mediae is rare, but the 
instances we have of this interchange are about equal in 
number in Ancient and Modern Greek. 

Then proceeding to the interchange of aspirates and tenues 
(if we leave out of consideration for the minute the com- 
binations a-(f), ad and a-^, on which so much stress is laid by those 
who maintain that % ^ were still real aspirates in classical 
times), we see that and t interchange as much now as then, 


and mostly in the dialects, whereas -y^ and k did so very 
frequently both in Attic and in the dialects, and still do so, and 
<f> and IT very seldom, but equally often in ancient and 

This interchange is extremely difficult to explain, especially 
as it still is of constant occurrence in Greece where % ^ are 
now, and have been for many centuries, distinct spirants. The 
interchange that is by far the easiest of explanation is that 
of K and T^ — the Greek spirant p^; is very difficult to pronounce, 
and as most foreigners, even Germans, find it impossible to 
manage, and say k instead, it is not hard to imagine that the 
people of some districts, e.g. Rhodes, find the same difficulty, and 
therefore regularly say k instead of X' ^^^ ^^^ ^^is peculiarity 
has in a few instances crept into general use. Similarly in 
Ancient Greek, whether ■)(^ was an aspirate or a spirant, its 
frequent interchange with k seems fairly easy to understand. 

T and 6 interchanged freely in the ancient language, but 
not more freely than they do in the modern, and the ancient 
forms with 6 for t can be paralleled by similar modern ones : 
thus for Attic 6p6(j)o^ we have the common modern forms 6pe<po) 
and dpo(j>ij and for Attic '^iOcov the modern cficodia for (ptoria 
and for Ionic aurt? the modern eiirv<; ; and modern ^d6paKo<s 
like ancient Ionic /3ddpaK0<;. So too for the interchanges of 
TT and ^ the ancient and modern languages furnish a fairly equal 
number of examples, e.g. like Attic -n-apdivoi; and (papOevo^ 
we have modern pd-n-avo'; and pd(j}avo<;, <pov')(Ta and vvktiti and 

If we then collect and weigh this evidence which proves 
that, though now assuredly spirants, the interchange between 
X^'^ ^"^^ their corresponding tenues still prevails, is it just to 
maintain that because these interchanges were so frequent in 


Ancient Greek, the explosive element in p^; ^ ^ must have still 
been distinctly audible, when, on the other hand, we find that 
these interchanges still continue and yet know that the explosive 
element is not heard in the present spirantic % </> ^ ; and correspond- 
ing forms with tenuis or spirant, such as ^ipra or ?ip6a (for rikda), 
are constantly met with and cause not the slightest diflSculty ? 

The question of course is, whether these forms with a tenuis 
are a remnant from olden times when % </> ^ were pronounced 
as true aspirates, or whether this interchange, still existent, 
could have been possible from the beginning even ii 'x,<f> had 
always been spirants. Or there is another solution which is 
held by some, and that is, that two pronunciations oi ■x^^ 
existed side by side in classical times — namely, that in some 
dialects they were as now spirants, and in others, in Ionic for 
example, real aspirates or, if not wholly and in all cases 
aspirates, yet generally so. In Ionic these aspirates constantly 
became tenues by the dropping of the aspirate, and in elisions 
the final tenuis when followed by a i-ough breathing, which 
in Ionic was hardly used, remained a tenuis. 

A survival of this aspiratic pronunciation is found in 
being pronounced as t by the present Asiatic Greeks, e.g. reXw 
for 0i\o). And we may perhaps go further and admit that in 
other dialects they were aspirates before or after certain letters, 
which is shown by their being frequently replaced by the cor- 
responding tenuis in such cases — in Elean, Boeotian and Lokrian^ 
may have been = t + h after cr, as in these dialects ar is generally 
found for a-9 ; so too in Crete the aspiratic pronunciation of 6 
was probably retained before p and v, as shown by the writings 
T/J, TV, though in other cases was in this dialect a spirant 
already before the fifth century B.C. (v. Meister ii. p. 54). 

As parallel to such a double co-existent pronunciation, we 


may perhaps adduce the English s, which as a rule is pro- 
nounced as a hard sibilant but in Somersetshire and Dorsetshire 
as a soft one, equivalent to z. 

The interchange of tenuis with % ^ ^ is so frequent in 
Modern Greek that it seems to entirely weaken the force of the 
argument urged by Curtiua, Meisterhans, &c., that this inter- 
change proves aspiratic proniinciation for the classical period. 
Because if we view this fact as an isolated one (apart from 
all other considerations which seem to point to 'x^cj) 6 being real 
aspirates), then " per se " can it be adduced to prove that % ^ ^ 
were aspirates, seeing that exactly the same interchange is 
found in the present language where X'f'^ ^^® pure spirants ? 
Why should, or rather how can, one and the same fact be taken 
to prove a certain thing in one instance which in another 
instance it cannot prove ? If the forms in the modern language 
which have a tenuis in place of aspirate are to be taken as 
survivals from a period when % ^ were = k + h &c. and 
hence easily convertible into tenues, we can say the same of 
similar classical forms, namely that they do not by any taeans 
prove aspiratic pronunciation oi x'P ^ ^'^^ ^'^^^ period, but are 
probably survivals of a period from 1000 — 2000 years anterior 
to the classical. 

Before we pass on to the interchange of aspirates with 
sibilants, we must say a few words about the few ancient 
examples of two consecutive aspirates being written for two 
consecutive tenues and which Curtius and G. Meyer take with 
the other interchanges of aspirates and tenues as evidence of 
the aspiratic nature of % ^. From inscriptions we get 
"Exffeop ( ="E«T«p) G.I.G. 7673 on a vase, and ej^^d? ( = 6«tos) 
I.G.A. 322&. 2 from Epidauros, and in Theocritus ii. 62 and vii. 
127 the form 67rt0^i5f« occurs for i-jniTTvo}. 


This last word has an exact parallel in the modern Cretan 
^9v(o for iTTvco and the vernacular <f)6epva for inepva, and all 
three have parallels in words from the poems of the middle 
ages, such as KparL-xOrjic^v for KparrjKTiKijv from Georgillas Jl. 
1490 A.D. and (j)Om^o\oyia (p. 54), ^iijidepoyvpevTdSe^ (p. 113), 
■X^pva-o^9epovyo<f)6pe and (f)pa,y(dr) from Wagner's Garmina Medii 
Aevi. The similarity of formation alone is a strong argument 
for a similarity of pronunciation of the letters in question, and 
it is heightened by the extreme unlikelihood of "E/erop being 
written "K'x^dop if it was to be spoken as "EKr-hop. The 
double aspirates seem the chief stumbling-block to the aspirate 
theory of % <^ ^, because it is more than difficult to imagine 
that if (j)d, yd, 66 were merely written through analogy to 7/3, 
Kir, "ITT, &c., but pronounced differently with the first aspirate a 
tenuis, we should not have countless errors on less carefully 
written inscriptions, such as vases, instead of only two or three 
e.g. utiOltov and KaTairdiiievrji;, neither of which is Attic by the 
way. And, as we have noticed in an earlier chapter, this juxta- 
position of two aspirates is in the most direct variance with 
the rules for the aspirates of the Indie language, so that the 
discovery of forms in the ancient language which have their 
exact counterparts in the language when % ^ ^ had beyond 
all dispute become spirants, tends strongly to convince us that 
already in classical times these letters were spirants. 

It remains therefore to notice the interchange of the aspirates 
with other letters. 

We will speak first of the interchanges of 6 and a. 

(a) 6 becomes <r chiefly in the Laconian dialect, for the 
stricter Dorians, as Merry and the old grammarians say, pre- 
served the 6 whilst the Laconians changed it into (t. Opinions 
differ as to the time when this change began. Ahrens places 


it as early as Alcman's time, and Blass apparently agrees with 
him ; G. Meyer thinks this is decidedly too early : he also 
gives as his opinion that the o- in our copies of the Lysistrata 
and Thucydides has been interpolated by later copyists ; Merry, 
on the contrary, takes the words of the Spartan herald as given 
in the Lysistrata 980 ff. as a specimen of the strict Laconian 
dialect. In these lines we have : 'Aaavav = 'Adrjvcov, (jLvari^at 
= /j,vdiaat, ai(o — 6ea>, opera = opdd, a^aeT6<; = dyado':, and 

In Thucyd. v. 77 tw aim tov a-vfji,aTo<; is Laconian for to3 0ew 
Tov Ovfiaro^, and in Arist. Mh. Nic. vii. 1 ceto? dvrip for ^eto? 
dvrjp. The old grammarians and Hesychius instance many 
other Laconian words written with an a for 6, such as icdaaet 
for Kade'i, irla-op for iriQo'i, &c. In our fragments of Alcman 
we have craXaa-aofiiSoicrav, a-d}<Xei, ecrrjKe, &c., but we also have 
6iaao<i, TrapdevoKuo, dvdo<s, Otolaiv, &c. 

On Laconian inscriptions o- is only found for 6 on those of late 
date, the earliest belonging to the second and first centuries B.C. 
and most of them to the time of the Eoman emperors. 
Amongst them are Set'ScKra? and SetVo/itTros of Hadrian's 
time G.I.G. 1241, Bm/xrea dveatjKev Cauer^ 34 of post-Christian 
time, TOV crlv ( = Qeov) ib. 33, probably of first century B.C. 
(v. Meyer, Greek Grammar § 211); ««? ar)pdTopiv ( = Kara 
d-qparopiov) ib. 36, 37 of the time of M. Aurelius. 

Blass says that the Laconian 6 was " undoubtedly for a time 
at least the Modern Greek spirant: if it had been a real s, 
it would have been so written by the Laconians themselves, 
while it is quite natural that the Athenians should have repre- 
sented the strange spirant by the allied sound a." This does 
not sound very convincing, as dialectical peculiarities of speech 
are never, or rarely, reproduced in the orthography of those 


born and bred to that dialect, but only by those who, for literary 
or other purposes, try to reproduce this peculiar pronunciation 
in writing. The Laconians may have spoken o- for spirant for 
a very long time before they themselves began to write it as o-, 
and Athenians of the time of Aristophanes would, if they pro- 
nounced 6 correctly as a surd spirant (granting for the moment 
that it was one), naturally in order to render literally the 
peculiar pronunciation of the Laconians, write an a- wherever 
the Laconians spoke 6 in that way. 

The 6 had therefore evidently become a spirant in Laconia 
at a very early date. 

If we now turn to other dialects we find that 6 becomes 
<T before fj. sometimes in Attic, e.g. in ending -a-fio for -d/io {v. 
Brugmann Crreek Grammar § 70, 1), thus ^a(rfj,6<{ is by Phryn. 
called the Attic form of the Ionic ^a6fi6<; — this distinction 
may be incorrect, but anyhow and o- are interchangeable in 
this word, ^aa-fio^ is found on a Mitylenian and a Lydian 
inscription (v. C.I.G. 2189 and 3486), dval3a0/jb6<! in Herod, ii. 
125, Kara^acr/jiO'; Aesch. Prom. 817. 

Similarly dvaK\avd/j,6<; and dvaK\avafi6<; Dion. H. vi. 46 
ed. Reiske ; pv0/i6<;, Attic, v. Lobeck. ad Phryn. p. 324, and Ionic 
jouo-yito? Archil, fr. 66 (Bergk p. 701), also in Anac. 78 ; and Call. 
Ef. 44, 5. IlaXao-tov for trcCKaQuov (Pax 574) and TpiKopvacoi; 
for -pv6io<; {Lys. 1032) may perhaps be taken as indications of 
a spirantic pronunciation of 6 in Attic. The Doric and Ionic 
eVXo? = ea-0\6<s (Find. 01. xii. 17). On an Elean inscription 
(Coll. 1172, 33) of probably the third century B.C. we have 
trorjao'crai = 'jroi'qa'aadai. 

Then again there is (&) the case of er becoming 6. The 
Rhodians, according to Strabo (xiii. p. 912), said ipvdl^r) for 
epva-c^rj ; in North looic Bv0fid<; stands for Sva-fMai; in Callim, 


H. ad Ger. 1. 10 ed. Meineke ; and on the Gortynian inscription 
66 or 6 is written for ad, e.g. ^prj6dai, airoBodat. 

All these cases of interchange between 6 and o- point strongly 
to a spirantic pronunciation of the former. 

In the modern language there seem to be no cases of (6), that 
is of (7 becoming 6. 

Of («), that is, of 6 becoming a, there are many examples 
to be taken from Locrian and Tsakonian. Thus in Locrian 
cr\iya) = 6\i^a), and aaXa/Movpa is a corrupt form of 66\o}/jLa. 
In Tsakonian crept = 6epo<;, a-eplvrov = depi^a, criva = ^t'?, 
a-aTr] = dvyaTTip, /cpia-d = Kpt6ij, &c ; especially to be remarked 
are a-ofio, cf. Laconian crepfj,oi, = 6€p/j,oi; Kaarjfiivoi — Ka6r}iievoi, 
cf. Hesych. Kaacrei, (Lacon.) : thus we have in Tsakonian, which 
is probably the direct descendant of Laconian, the survival of 
the ancient Laconian pronunciation of 6 as a-. 

In the contemporary language we see that p^;, as well as 6, 
becomes o- occasionally. In Cypriot this change is very general, 
thus, a-ept = %e/3f, creiXr), a'qpa, eroipo<;, &c., for j^eiXr], XVP^' 
'Xplpo'i. A similar pronunciation prevails in Amorgos, Kalymnos 
and Astypaleia, e.g. ecrei = e'^ei, i^oa-ij = e^o')(rj, &c. Likewise 
in Samothrace and in the Pontic dialects, e.g. Trebizond, there 
are the forms craipofj-ai = j^aipofiat, creKihcov = •^eXiBcov, &c. 
In Calabria creifi&va — ^(eifimp, aelpo = 'xelpov ; in Macedonian 
aoivUr) == ^oii/tf (also in Pontic we have ■)(^oiviKov and o-oIvikov, 
and in Tsakonian y^^otviKo). 

We do not know of any extant instances of this change in 
the ancient language. 




This subject has such an important bearing on the probable 
pronunciation o{ '^<p in the various dialects that we must first 
briefly review the different ways in which it is treated in them 
(our limited space will not allow of our entering into full details), 
and then see what conclusions are to be drawn therefrom as to 
the nature of % ^ ^. 

In Attic, both old and new, the rule is that the final tenuis 
of a word is changed into its corresponding aspirate before an 
initial " spiritus asper," e.g. xad' rifiepav, ixp' ev6<;. 

And this rule is never (except for a few exceptions on G.I. A. 
i. 324, which is generally acknowledged to be the work of an Ionian) 
transgressed on any inscription before the Christian era, in spite 
of the fluctuating use of the spiritus asper which is often omitted, 
when initial, on the older inscriptions and has no sign to repre- 
sent it at all after the formal adoption of the Ionic alphabet. 

The orthography which corresponds exactly to the pronunciation 
is that of words like ecfyir/fu and that of those inscriptions where 
the initial spiritus asper is not written after the aspirate, e.g. 
Kadd CIA. iv. 61 a. 26. Writings like KaOhdirep on same in- 
scription, and d^' oii, d(j}' lttttov of the Alexandrian grammarians 
are incorrect, though the latter may, if had already become 
a spirant in their time, be explained by their not knowing that 
after a spirant the rough breathing ought, as after an aspirate, 
not to be pronounced. 

In Asiatic Ionian, where the rough breathing was lacking, 
we have complete psilosis on the inscriptions except for a few 


cases like fiedeXj) B. 174 a, Kadrjfievov ih. 156, KaOohov ih. 238, 
found on inscriptions of the 4th or 5th century B.C., which are 
probably old compounds in which the 6 has persisted, and for 
others like KaOd-Trep B. 147, 151, Kadia-Tafievo^ B. 158 (with 
dTnjyyjaiv) which are probably due to Hellenistic influence. 

In Asiatic Aeolian we have no inscription from the time when 
the sign of the "spiritus asper" was still in use which has 
instances of elision or crasis, and on all the others there is 
psilosis, and the same applies to the evidence from MSS. 
Meister considers that Bergk rightly rejects as incorrect the 
signs of aspiration in the fragments of the Lesbian poets 
(Sappho, Alcman) ; of Alcman more hereafter. 

Exceptions to this rule occur on inscriptions from the time of 
Alexander onwards, e.g. d(j)iK6fjLevo<s Coll. 281, 15, Kadd ih. 311, 9, 
i(f)' av, €((>' olcriv ih. 311 must again be 'old compounds' or due 
to Hellenistic influence. 

In Boeotian, aspiration of the tenuis is regularly observed, e.g. 
■n-oeelXero Coll. 488, 122. 

In the dialects of N. W. Greece, e.g. Phokian, Aetolian, aspiration 
of the tenuis is generally observed except in Lohrian where the 
tenuis is never aspirated in cases of elision and crasis on the older 
inscriptions, but only on two of later date (Coll. 1502 and 1508). 

In Mean, psilosis obtains except for a few instances similar to, 
and probably due to, the same causes as those noticed under 
Ionian and Aeolian. 

In Doric, which must be subdivided, we find 

(a) in Laconian, aspiration of the tenues on inscriptions is 
the rule, e.g. iroO' dfii Newton ii. 143, Ka0' a O.I.G. 1346; 
KaT Ihiav and «a^' ihiav represent two different forms of tSto?, 

Apart from the inscriptions, Apollonius testifies that in Doric 
poets, among whom Alcman is evidently to be included, the 


tenuis in elision and crasis is constantly not altered before the 
spiritus asper, e.g. kco to^otu^ KoXXoaT viravkev {v. Blass, p. 112), 
and of the fragments he gives Bergk attributes some to Alcman. 
As there is great uncertainty not only as to what is Alcman's, 
but also as to the correct reading in the extant fragments 
{■)(WTrdpav fr. 76 seems certain), it is difficult to draw any 
definite inference with regard to the treatment of the tenues in 
crasis and elision. Blass, in suggesting that the reason of this 
non-aspiration may be found 'va.-)(^<f>6 being in Laconian spirants 
already, seems entirely to leave out of account the inscriptions. 

(6) in Tarentum and Heraldeia, aspiration of the tenues takes 
place, so also in Messenian and Goriwthian inscriptions. 

(c) in Bhodes also, aspiration is regularly observed except in 
the word tepo?, which is in this dialect written tepo?, and in 
KarvirepOe I.G.A. 482 a. 

(d) in Crete psilosis prevailed and on the Gortynian inscrip- 
tions the spiritus asper is never written, and r is never changed 
into d. 

To sum up, in Attic, Boeotian, Rhodian, the Doric of Tarentum, 
Herakleia and perhaps Laconia, aspiration of the tenues in 
elision and crasis is roughly speaking uniformly observed, 
whereas it is neglected by the Asiatic Aeolians and lonians, the 
Eleans and the Cretans. 

In the latter set of countries the initial spiritus asper had 
been entirely lost before historic tiimes, and consequently there 
was nothing to cause aspiration of a final tenuis. 

The regular aspiration of tenues in Attica, &c., proves accord- 
ing to some the true aspiratic nature of x4'^- According to 
Von der Miihll, if there was any reason whatever for the ortho- 
graphy ■x^cj) in elisions, it can only be found in the supposition 
that the pronunciation of % ^ was identical with that of the 



respective tenuis + spiritus asper. For, if this did not represent 
the pronunciation of %^^, what led to the substitution in 
question ever being introduced ? 

On the other hand, if we suppose % ^ ^ always to have been 
spirants in Greek, we must conceive that in the case of a tenuis 
followed by an initial spiritus asper the two coalesced into the 
ne sound of a spirant. 

It might be argued on the other side that the tenuis + 
spiritus asper must by the time of the introduction of (j) and y( have 
already changed more or less from real aspirates, for otherwise 
how is it possible to explain that never once in Attica before the 
time of Eukleides whilst H still = h, nor in any other dialect, as 
for instance in Herculaneum where they had a special sign |- for 
" h," do we find on an inscription of any kind ttA, kA, or rh 
written in elisions, erases and compound words for ip, '^ or 6 
(excepting of course the cases of Thera and Melos) ? 

On Attic inscriptions aspirates and tenues are freely inter- 
changed, the " h " is often left out and sometimes wrongly inserted, 
we have mistakes in compounds and elisions, e.g. d^earaXKa, 
irevr opKiav, we even have KaOhdirep near KaOd G.I.A. iv. 61 a 26, 
but nowhere have we any error such as kut hefiepav, KaThiarTacn 
or dirheaTaXKa, which, if ^ = tt + A &c., are just the very ones 
to be expected on the more inaccurate and earlier inscriptions. 
Similarly on the tables of Herakleia we consistently find tto^oSo?, 
TTodeXofjievo';, &c., not once ttotI-oSo?, or ■7roT\-eXofjievo<i. Such 
mistakes as a<f)iaTa\Ka are plainly enough due to the stone- 
mason's ignorance as to whether earaXKu was aspirated or not, 
and is like the Mod. jxedavpiov for /Meravptov, e^eros for eTrero?. 

The writings KaOhd'n-ep, dtfihov &c. are nonsense, if 0, <j) of 
themselves sounded as r-\-h,Tr+ h; but if they were spirants, 
the stonemason who knew by ear that KaOdirep not Kard-Trep 


was said, and further knew dimly that direp by itself was an 
aspirated word, thought he had better put in its " h," not under- 
standing that its aspirate was absorbed by the spirant 0. At all 
events ii xi'^ were real aspirates, then we ought, if only in 
elisions, to have some instances of «A, irh, rh being written for 
them before the abolition of the sign of the spiritus asper, and, 
this not being the case, it is extremely difficult to believe that 
X<f>0 were still real aspirates at the time from which even our 
earliest Attic inscriptions date. 

It is also maintained by some, e.g. Schiltz, that the " spiritus 
asper " had disappeared in Attic by the fifth or fourth century 
B.C., and others admit that, although it was in their opinion still 
heard down to at least the beginning of the Christian era, it had a 
very weak sound. If this were so, are we not almost forced to 
believe that the tenuis and spiritus asper must have coalesced 
into one sound before the latter became considerably weak- 
ened, and certainly before it quite disappeared ? 

Again, although the writing (f> and ;)^ for tenuis and spiritus 
asper could not possibly have come into use before the introduc- 
tion of these signs, we are not in a position to state that the 
language was not ready for them before — the tt + A and K + h 
may have become affricatives or spirants some time before single 
signs were introduced to represent them. 

In post-classical times there are an increasing number of 
exceptions to this rule of aspiration and these are either to be 
explained by % </> ^ now more and more approximating to 
spirants, and their consequently no longer correctly representing 
a tenuis + "spiritus asper," or by a general ignorance of 
orthography at a time when the initial " spiritus asper " was 
probably completely lost to the language, and further by 
some popular forms in which aspiration existed, though not 


sanctioned by tradition, having come into general use, e.g. eTot 
for eVos. 

As in modern Greek Kadox;, a<^' orov (or vulg. a^ovia), 
a(f>ir]ij,i, Kadrjfiipav are said, but air' on and a-jr oirov, because 
the former are stereotyped forms which have come down from 
antiquity, whereas the latter did not exist in the ancient 
lanofuaoe but have been formed in latter times when the 
spiritus asper was no longer spoken, so too in Attica and else- 
where in post-classical times we find non-aspiration of the 
tenues in uncommon elisions, whereas it is always retained in 
old forms and elisions which by their frequency had come to be 
regarded as almost one word. 

Now if we turn to the dialects of Asia Minor, Elis, &c., we 
find psilosis, caused undoubtedly by the spiritus asper having 
disappeared in these dialects before it and the preceding final 
tenuis in compounds and elisions had had time to coalesce into 
one sound, and thus we get air ov (for ^i^' ov) I.G.A. 246, rr/jOiyt 
(=T^"Hp27) B.211, &c. 

But even in these dialects we get occasional examples of 
aspiration such as d<f)iK6fievo<; Coll. 281, 15, KadoerTo-fj-evoi;, Kadwp, 
/j,e6e\,Tj B. 174 a, KudevSa), which correspond exactly to the tradi- 
tional forms preserved in Medieval and Modern Greek and, like 
them, are to be explained as old compounds dating from a time 
when the spiritus asper still existed in these dialects and had 
grown so firmly into one sound with the preceding tenuis, that 
these forms were not affected by the loss of the spiritus asper — 
or they may be explained by Attic influence. The explanation of 
analogy does not seem a good one, as the examples occur in 
different places and times and on different inscriptions. Other 
exceptions there are too in these dialects, which, according to 
Meister, are due to Hellenistic influence. 


The word KaTvirep6e I.G.A. 482 on the Abu-Simbel inscrip- 
tion from Nubia, which is probably written by a Ehodian, is 
noticeable for the non-aspiration of the t, because on the same 
inscription the " h " is written elsewhere, e.g. ho, but not 
regularly by any means. It can be explained in two ways ; 
either, if Q is supposed = t + /i, it was caused by dissimilation of 
the aspirates, or else 6 was a spirant. 

The Attic forms ' KtrrfK.imTri'i C.I.G. 518 and KpaTiir-n-ot; 
Coll. 1307 (c/. 'A(})r]\ia,Tri<; C.I.G. 6180, 6181) cannot be taken 
as instances of the neglect of aspiration, for it must be remem- 
bered that no rule can be postulated for the erratic treatment of 
proper names. As regards KpdTnnro<; we^ find that the word 
tTTTTo? was regularly in all dialects deprived of its rough breath- 
ing when it formed the second member of a compound, thus 
we have Boeot. 'AXKiiriroi;, Delph. TLpdriinro^, &c., which is 
explained by the aspirate in i'tttto? not being Indo-Germ. ; 
whilst rjXto? when in composition was sometimes aspirated and 
sometimes not. 

To conclude, in those dialects where regular aspiration of a 
final tenuis before a following " spiritus asper " takes place, it 
seems only natural to assume that the ^ </> or ^ in these 
elisions must have been true aspirates, but on deeper investiga- 
tion we see that one argument against this conclusion is the 
fact that we have no errors on inscriptions of tenuis + " spiritus 
asper " being written instead of a 'x^^ ot 6 ; and, secondly, that 
if the spiritus asper had grown so weak as ,no longer even to 
require a distinctive sign, it was probably no longer of sufficient 
force to retain its influence in every case over a preceding 



Although in the course of the preceding pages we have 
from time to time drawn conclusions from the various arguments, 
it may be useful now to summarize. 

In the first place it will be as well to say at once that we 
consider the question one that does not admit of any definite 
solution, because even the safest, viz. the internal evidence of 
the language itself, is both of an uncertain and a conflicting 
nature. This being so, we can, after carefully sifting the same, 
do nothing beyond forming a more or less certain hypothesis 
from estimating the value of the arguments on either side and 
trying to duly appreciate them. From such an estimate we 
obtain the following results. 

In support of the aspiratic theory, we have the two analogical 
phonetic laws in Sanskrit and Greek, by which two consecutive 
syllables cannot begin with an aspirate, and a root may not 
begin and end with an aspirate. Add to this the a priori 
evidence found in the process of elision, and we have the main 
arguments for the aspiratic theory. 

On the other hand, in support of the spirantic theory, we have 
the difference of phonetic law in Sanskrit and Greek, by which 
in the latter language we find combinations of aspirates. As 
regards internal evidence, with the exception of that furnished 
by elision, it would seem to favour this theory. That it does 
so, we have attempted to show in our investigation of the evolu- 
tion of the phonetic laws and the history of interchange, which 
in our opinion seems to point to a continuity of pronunciation. 

As to the testimony of the grammarians, we think we have 


shown by our exposition that, if considered impartially and in 
its entirety, it cannot be looked upon as reliable evidence for 
either theory. 

These are the broad conclusions at which we arrive, and we 
do not think they are such as to justify a final decision in favour 
of the two opposed theories which wo have attempted to